Intravehicular LLM-based stress management

Abstract

A system includes a processor and a non-transitory computer readable medium. The non-transitory computer readable medium includes programming instructions that when executed by the processor, cause the processor to operate a large language model, and generate a prompt chain using an emotional state information providing an emotional state of a driver. The prompt chain includes a sequence of parameterized prompts, where the sequence of parameterized prompts includes a first parameterized prompt that incorporates information gleaned from a response by the large language model to a second parameterized prompt that occurred prior to the first parameterized prompt. The programming instructions further cause the processor to output, by the large language model, a candidate action to be undertaken to regulate the emotional state of the driver based the prompt chain and a context information, and cause the candidate action to be performed in response to receiving approval from the driver.

Description

INTRA VEHICULAR LLM-BASED STRESS MANAGEMENTCROSS-REFERENCE TO RELATED APPLICATIONS[00011 This application claims the benefit of U.S. provisional application Serial No. 63 / 659,000 filed June 12, 2024, the disclosure of which is hereby incorporated in its entirety by reference herein.TECHNICAL FIELD

[0002] The present disclosure is generally directed to a system and / or method for controlling a state of a driver of a vehicle.BACKGROUND|'00031 Until such time as autonomous vehicles become widely adopted, humans must continue to drive cars. Since the motor car was invented over a century ago, driving can be a novel and exhilarating experience, but can also be one of the more stressful activities that humans endure on a daily basis.

[0004] The stress associated with driving can arise from traffic, weather, and hazardous or emotionally provocative behavior from other drivers.

[0005] All of this takes a toll on the driver’s emotional state. The sheer stress of driving, combined with other stress factors in one’s life, such as a high workload, family responsibilities, and various distractions, such as hunger and thirst, combine to create a perfect storm that can affect the driver’s emotional well-being.

[0006] In some instances, the driver may lapse into an emotional state that may not be conducive to safe driving. For example, cortical activity can fall, attention is tunneled, and situational awareness declines, all of which can impair the driver’s ability to anticipate events and to ignore otherwise distracting stimuli.SUMMARY

[0007] In some aspects, the present disclosure is directed to a system for regulating an emotional state of a driver of a vehicle. The system includes a processor and a non-transitory computer readable medium comprising programming instructions that when executed by the processor, cause the processor to operate a large language model and generate a prompt chain using an emotional state information providing the emotional state of the driver. The prompt chain includes a sequence of parameterized prompts, where the sequence of parameterized prompts includes a first parameterized prompt that incorporates information gleaned from a response by the large language model to a second parameterized prompt that occurred prior to the first parameterized prompt. The programming instructions further cause the processor to output, by the large language model, a candidate action to be undertaken to regulate the emotional state of the driver based the prompt chain and a context information, and cause the candidate action to be performed in response to receiving approval from the driver.

[0008] In other aspects, the present disclosure is directed to a method for regulating an emotional state of a driver of a vehicle. The method includes generating a prompt chain using an emotional state information providing the emotional state of the driver. The prompt chain includes a sequence of parameterized prompts, where the sequence of parameterized prompts includes a first parameterized prompt that incorporates information gleaned from a response by a large language model to a second parameterized prompt that occurred prior to the first parameterized prompt. The method further includes outputting, by the large language model, a candidate action to be undertaken to regulate the emotional state of the driver using the prompt chain and a context information, and causing the candidate action to be performed in response to receiving approval from the driver.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] FIG. 1 illustrates an example empathetic agent implemented in a vehicle.

[0001] FIG. 2 illustrates an accretive parameterized prompt chain implemented by the empathetic agent of FIG. 1.

[0011] FIG. 3 illustrates an example accretive prompt sequence.DETAILED DESCRIPTION

[0012] The following description is merely exemplary in nature and is not intended to limit the present disclosure, application, or uses. It should be understood that throughout the drawings, corresponding reference numerals indicate like or corresponding parts and features. The figures are not necessarily to scale; some features may be exaggerated or minimized to show details of particular components.

[0013] In one aspect, the present disclosure is directed to an empathetic assistant system, or empathetic agent, for promoting road safety by reducing the likelihood that a trigger-induced stress change in a driver’s emotional state may lead to unsafe driving and improving the driver’s overall driving experience.[0014| The empathetic agent is implemented, at least in part, by a large language model (LLM) agent that receives driver-specific information and situational context information. Examples of driver-specific information include transient information, such as the driver’s emotional state, and static information, such as known state-change triggers. To collect the driver’s emotional state, the empathetic agent uses an emotion-detection engine or emotion detector module that acquires information such as the driver’s voice, the driver’s image, and information from wearable devices, such as smart watches, smart phones, and / or health monitors (e.g., FITBIT™). The emotion detector uses that information to assess the driver’s emotional state. That is the emotion detector detects whether the emotional state of the driver is at an optimal emotional state for driving the vehicle. In a non-limiting example, the optimal emotional state may be quantified by defining a target heart rate, voice inflections, and / or facial characteristics.

[0015] Examples of situational context information include current traffic information, weather information, and vehicle information. Also included within the scope of situational context is information that is personal to the driver, such as information about calendared events and other temporal constraints; driver preferences; and statistics about the driver’s habits.

[0016] The empathetic agent acquires situational context from such sources as real-time traffic applications, which provide information about impediments to smooth traffic flow, such as automobile accidents, heavy traffic, road construction, and lane closures.|0017] Additional situational context arises from sources such as a user profile or driver preferences and a calendar. The calendar provides such useful information as appointments, meetings, and tasks that the driver is planning to carry out. The user profile provides information such as preferences recorded while the driver is operating the vehicle and a history of the driver’s interaction with the agent.

[0018] Triggers are events that are likely to cause a change in the driver’s emotional state. The empathetic agent is configured to analyze the triggers to assess relevance thereof and to autonomously select, based on all context cues, an action that has been selected to cause the driver’ s emotional state to resist changing in response to exposure to a trigger or to avoid exposure to the trigger altogether. Such an action is referred to herein as a “trigger ameliorator.”

[0019] Examples of a trigger-ameliorator include actions that simply circumvent the trigger. An example would be a case in which a known trigger is heavy traffic and situational context indicates traffic along the present route. Under such circumstances, the LLM agent would autonomously prepare a new route to avoid the heavy traffic, and hence the trigger. The agent would then announce this intent to the driver, explain its purpose, and invite approval. As a result of this foresight on the part of the agent, the driver has an opportunity to avoid being exposed to the trigger.|0020] Other examples of trigger-ameliorators include those that attempt to mitigate the impact of the driver being exposed to the trigger. In one example, upon recognizing that there is no practical way to reroute to avoid heavy traffic, the empathetic agent autonomously selects a deepbreathing exercise that is tuned to distress the driver’s emotional state. The empathetic agent notifies the driver of the triggering event (e.g., heavy traffic), and prompts the driver of an available breathing exercise to be guided by the empathetic agent to possibly reduce the stress level of the driver.

[0021] The empathetic agent is also equipped with a feedback loop to allow it to continually assess whether the autonomously selected action is effective at mitigating the impact of the trigger. As a result, the empathetic agent is able to select an alternative trigger-combatting action if a selected acted is not effective. For example, if after several minutes of deep breathing the empathetic agent observes that the trigger has begun to adversely affect the driver anyway, the empathetic agent, in recognition of music’s role in human mood control, may recognize that the driver recently played music from a particular playlist or genre during heavy rush hour traffic. The empathetic agent, having realized that the deep breathing is not working, may then inquire whether the driver would like to listen to a song from the playlist or genre Based on the driver’s answer, the empathetic agent will begin to learn the driver’s preferences. As a result, the empathetic agent grows more effective over time.

[0022] Yet other examples include in-car control modifications. For example, if the situational context indicates that prevailing winds are steering smoke from a nearby forest fire towards the section of road that the vehicle is about to traverse, the empathetic agent notifies the driver of the fire. The empathetic agent may also ask the driver for permission to roll up the windows and to set the climate control system to recirculate air in the cabin. In response to approval of such a request, the empathetic agent proceeds to do so, thereby reducing the likelihood of an emotional state change as a result of having been exposed to smoke from a forest fire.10023] The empathetic agent is implemented as an LLM agent with chain-of-thought prompting. The empathetic agent makes use of a parameterized prompt template to analyze the situational context for the existence of triggers and to reason on a step-by-step chain of intermediate reasoning steps, e.g., a “chain of thought,” to arrive at a suitable trigger amelioration action. An example of a thought-chaining framework is LANGCHAIN™-type framework.

[0024] The empathetic agent makes use of various external tools to obtain situational context and calls upon external applications to implement its autonomously selected actions. The empathetic agent thus acts proactively without the need for an explicit user request. This is particularly advantageous because during times of high emotional stress, the last thing on a driver’s mind is often the act of interacting with the empathetic agent. As a result, it is possible to implement theempathetic agent without the effort of conventional natural-language understanding development and localization. All that is required is to provide plugin descriptions and to include, in the prompt, a parameter that identifies the desired language.10025] In another aspect, the features of a system for regulating an emotional state of a driver of a vehicle so that the emotional state is an optimal driving- state. This optimal driving-state is one that is optimal for a driver to be in while the driver is operating the vehicle. Such a system includes a LLM and an empathetic agent. The empathetic agent receives emotional state information, which is indicative of the emotional state, and context. It also provides a prompt chain to the model. This prompt chain includes a sequence of parameterized prompts, among which is a first parameterized prompt that incorporates information gleaned from a response by the model to a second parameterized prompt that occurred prior to the first parameterized prompt. The empathetic agent is also configured to receive, from the model, a recommended action to be undertaken to regulate the driver’s emotional state given the external context, to offer this recommended action to the driver, and, upon the driver’s approval, to cause this recommended action to be realized.

[0026] Some embodiments also include an emotion detector that uses the information that is received from a cabin monitor to generate the emotional state information.

[0027] Still other embodiments include those in which the context includes one or more of: weather information from a weather service, traffic information from a traffic service, calendar information from a calendar service that provides information concerning an appointment, and driver preferences.

[0028] Also, among the embodiments are those that include a cabin monitor that provides information to the agent. Such a cabin monitor includes a camera, a microphone, and a physiological sensor.

[0029] In still other embodiments, the empathetic agent forms an aggregate of the emotional state information and the context and provides the aggregate to the model.

[0030] As provided above, the trigger-ameliorator, which may also be referred to as a candidate action, is selected to reduce the likelihood of exposing the driver to a trigger that urges theemotional state away from the optimal driving-state. The candidate action may include, but is not limited to, rerouting the vehicle such that it traverses a second route that differs from a first route along which the vehicle was previously routed. This is carried out when traversal of the first route would expose the driver to a trigger that would be avoided by traversing the second route instead of the first route. The trigger in this case is one that urges the emotional state of the driver away from the optimal driving-state.

[0031] Other examples of the candidate action include controlling a feature of the vehicle in a manner that has been selected to urge the emotional state of the driver towards the optimal drivingstate, such as having an infotainment system in the vehicle play music that has been selected to urge the emotional state of the driver away from a sub-optimal driving state. The music is selected based on driver preferences that are available to the empathetic agent. In yet another example, the candidate action includes coaching the driver to carry out an action (e g., breathing exercises) that urges the emotional state of the driver to remain in the optimal driving-state.

[0032] FIG. 1 shows a vehicle 10 that is being driven by a driver 12. The driver 12, being a human, will have a time-varying emotional state 14 which is illustrated as a clock. This emotional state 14 changes in response to certain triggers 16. Among these emotional states 14 is an optimal driving-state 18 and one or more sub-optimal driving- states 20.

[0033] In some aspects, the vehicle 10 includes an infotainment system 22 that connects to a cabin monitor 24. The cabin monitor 24 includes a camera 26, a microphone 28, and a physiological sensor 30 disposed within the vehicle’s cabin. The camera 26, microphone 28, and the sensor 30 of the cabin monitor 24 are positioned so as to observe features that are indicative of the driver’s emotional state 14. For example, the camera 26 is configured to observe the driver’s countenance, the direction of the driver’s gaze, and any gestures made by the driver. The microphone 28 is configured to detect the tenor of the driver’s speech. The physiological sensor 30 is configured to measure parameters that are known to be connected with emotional state 14, such as heartrate, skin conductivity, which is tied to levels of perspiration, and gripping force on the vehicle’s steering wheel. The physiological sensor 30 may use different sensory devices (e.g., heart rate sensor, pressure sensor on steering wheel, among others).

[0034] The infotainment system 22 comprises a processing system that executes various applications. In some embodiments, the processing system is local to the vehicle 10, whereas in other embodiments, the processing system is distributed and thus features a local component and a remote component. In some aspects, the infotainment system 22 is configured to include an emotion detector module 32 that receives information from the cabin monitor 24 and uses it to estimate the driver’s current emotional state 14. The emotion detector module 32 thus outputs the driver’s emotional state information 34.

[0035] In some aspects, the emotion detector module 32 is configured to use one or more cues for assessing the driver’s emotional state 14. These include observation of sentiment cues arising from speech, including content and prosody thereof, as acquired by the speech interface from the microphone 28, facial expression and gaze, as acquired from the camera 26, heart rate, breathing rate, and body temperature, as acquired from the physiological sensor 30. Other cues include observation of time-varying kinematic parameters, including, for example, velocity, acceleration, and jerk.|0036| The infotainment system 22 may also be configured to include is a speech interface module 36 that includes an automatic-speech recognition module (ASR) 38 and a text- to-speech module (TTS) 39. The speech interface module 36 is thus useful for providing additional linguistic cues that can be beneficial for modelling the semantic context 42. Examples of such linguistic cues include how fast the driver 12 is speaking and how loud the driver 12 is speaking.10037] The infotainment system 22 also calls on various external services 44 through the use of respective application-program interfaces. Among such services are a weather service 46, a traffic service 48, and a calendar service 50 that has access to the driver’s appointments and errand list. These external services 44 are useful for providing external context 52. In a non-limiting example, the infotainment system 22 may receive information via connection a smart phone that is in communication with the infotainment system 22 providing access to software applications on the smart phone. In another example, the infotainment system 22 receives the information from institutions associated with the external services 44 that transmit the information via a wireless communication supported by a vehicle-to everything communication feature of the vehicle 10.

[0038] The infotainment system 22 also has access to various driver preferences 54. These are obtained directly from the driver during a registration process or learned by observation. The combination of the emotional state information 34 and the driver preferences 54 defines a driver’s “persona.” Examples of information that form part of the driver’s persona include entertainment preferences, such as preferred forms of music and preferred software applications for execution (e g., music platforms, podcast applications, etc ). These are typically found in the driver preferences 54.

[0039] Other examples of such information include common triggers 16. As used herein, a “trigger” is a set of circumstances or one or more events that increases the likelihood of a transition from the driver’s optimal driving-state 18 to the driver’s sub-optimal driving-state 20. Another name for a “trigger” is thus a “stressor.”

[0040] Triggers 16 are generally personal to the driver 12. They range from conventional ones, such as sensitive to traffic jams, high workloads, or driving in heavy snow, to more unusual ones, such as fear of driving adjacent to large trucks and buses or discomfort at the sound of the bells of an ice-cream truck. Information about the existence of such triggers 16 is available from monitoring the traffic service 48, the driver’s calendar service 50. and the weather service 46, respectively.

[0041] Still other examples of information that forms part of a driver’s persona include the driver’s preferences for the use of various features of the vehicle 10, such as its navigation tools, advanced driver-assistant features, climate-control features, and entertainment features.

[0042] Still other examples of information that forms part of a driver’s persona include information characteristic of driving context, such as statistics concerning distances and times spent traveling, including both time durations and times of day. This too is found in the driver preferences 54.

[0043] Also among the examples of information are the driver’s preferred solutions to various problems. For instance, a learning feature may recognize that a particular driver 12 prefers to replenish the on-board energy supply of the vehicle 10 long before projected exhaustion of thaton-board energy supply. This applies to on-board supplies of chemical energy, such as gasoline, or on-board supplies of electrical energy, such as electrical charge or reagents for fuel cells. Such information is likewise found in the driver preferences 54.

[0044] In some aspects the driver preferences 54 may be stored in a memory device of the infotainment system 22.

[0045] An empathetic agent module 56 that is in communication with a model module 58 (i.e., model 58) executes on the infotainment system 22. In a preferred embodiment, the model module 58 is a large language model. While illustrated separately, the LLM module 58 may be provided with the empathetic agent module 56 in some implementations. The empathetic agent module 56 and the LLM module 58 may be implemented in various suitable ways using a processor and a non- transitory computer readable medium comprising programming instructions to perform the operations described herein when executed by the processor.

[0046] The empathetic agent module 56 is an Al system that uses the model module 58 to understand and generate a simulacrum of natural language human communication, via text or speech, and to select one or more candidate actions in an effort to reach a defined goal. The empathetic agent module 56 connects to one or more tools and / or application program interfaces to obtain information to assess a situation and to propose an appropriate response to the situation. The empathetic agent module 56 is directed through prompts that encode both the persona and instructions. In a non-limiting example, the empathetic agent module 56 is configured using a LANGCHAIN™- type of framework using LLMs in combination with a generative Al system (e.g. , CHATGPT™- type of generative Al).

[0047] The empathetic agent module 56 receives emotional state information 34, the linguistic context 42, the external context 52, and the driver preferences 54. Based on the amalgamation of this information, the empathetic agent module 56 provides a prompt chain 60 to the model module 58 and outputs an action signal 62, which is indicative of one or more candidate actions. The action signal 62 identifies one or more candidate actions, if carried out, would urge the driver’s emotional state 14 to remain or move towards the optimal driving-state 18. Examples of actions having this property include turning on selected music 64, controlling a component of the vehicle 10, rerouting68 the vehicle 10, or using the text-to-speech module 39 and the loudspeaker 72 to provide coaching instructions 74 to coach the driver through a mental exercise 70. In response to receiving a selection from the driver 12, the selected candidate action is performed. In some cases, the driver 12 selects no action at all and instead endures whatever stress may arise.

[0048] In some aspects, the infotainment system 22 includes user interfaces disposed in the vehicle 10 to communicate with the driver 12. The user interface may include but is not limited to a touchscreen display, a speaker and microphone, and / or buttons.

[0049] Because the empathetic agent module 56 has access to real-time emotional state information 34, it is able to observe the effectiveness of whatever action in the action signal 62 is ultimately realized. This creates a feedback loop. As a result, the empathetic agent module 56 acts as an emotional state regulator that uses feedback control to promote the likelihood of the vehicle 10 being operated while its driver 12 is in the optimal driving- state 18.

[0050] Referring now to FIG. 2, the empathetic agent module 56 collects information about the driver’s persona, such as semantic context 42, driver preferences 54, and emotional state information 34 and information about external context 52 to the empathetic agent module 56 .

[0051] The empathetic agent module 56 then provides a first parameterized prompt 75 to the model module 58. The model module 58 then provides first response 76 back to the empathetic agent module 56 . The empathetic agent module 56 then uses this first response 76 as a basis for formulating a second parameterized prompt 78, which it then provides to the model module 58. This second parameterized prompt 78 includes information gleaned from the first response 76.

[0052] As the foregoing procedure continues, the result is the accretive prompt chain 60 in which each parameterized prompt 75, 76 amounts to a reasoning step. Each parameterized prompt 78, except for the first prompt 75, includes an accretion of results from preceding reasoning steps. The first parameterized prompt 75 is based on information available to the empathetic agent module 56.

[0053] At the end of the sequential chain, the empathetic agent module 56 ranks the triggers 16 and identifies the most relevant actions 62 to relieve or avoid stress from those triggers 16. FIG.3 shows an example of a tool routing and function binding operations 100 using LANGCHAIN™- type of framework for assessing several different options in addition to those already discussed, such as introducing fragrance and telling a joke.

[0054] The empathetic agent module 56 then presents one or more actions 62 to the driver 12. It does so using NLU-like annotation, thus enabling the re-use of an existing integration layer.

[0055] Once a candidate action has been agreed to by the driver 12, it must actually be carried out. This may be done using an API call to an existing application if one is available. For example, if the action 62 approved by the driver 12 involves mediation for stress reduction, the agent 62 causes an API call to be made to a commercially-available software application (e.g., Headspace™), which then takes over from the empathetic agent module 56 .

[0056] The empathetic agent module 56 relies on a variety of criteria to select an appropriate action 62. In some cases, the empathetic agent module 56 selects an action 62 to pre-emptively avoid exposing the driver 12 to known triggers 16, for example by re-routing to a route having a lower stress index.

[0057] In other cases, the empathetic agent module 56 selects an action 62 that modifies an existing situation in an effort to blunt the effect of any triggers 16 arising from that situation. This can be achieved, for example, by modifying the vehicle’s operation. One way to do this is to activate the vehicle’s advanced driving assistance system.

[0058] In still other cases, the empathetic agent module 56 selects an action 62 that redirects the driver’s attention away from the trigger 16, thus reducing that trigger’s effect on the driver’s emotional state. This can include adjusting lighting within the vehicle’s cabin, injecting fragrance into the vehicle’s cabin, selecting music based on the circumstances and playing that music, or using the climate control system to adjust temperature and humidity of air within the cabin.

[0059] Since the empathetic agent module 56 monitors the driver’s emotional state, any action that is ultimately realized will be subject to feedback control by the empathetic agent module 56 . Thus, selecting and playing music based on circumstances, when coupled with observation of the driver’s emotional state amounts to musical biofeedback.

[0060] In some cases, the empathetic agent module 56 instructs or causes the driver 12 to be coached in such a way as to ameliorate the physiological effects of the trigger 16, for example by breathing or mediation exercises.

[0061] As a result of its activity, the empathetic agent module 56 assists the driver 12 in maintaining a state-of-mind conductive to safe driving. In doing so, the empathetic agent module 56 improves safety for all who use the public roads, whether drivers or pedestrians, while concurrently benefiting the driver 12.

[0062] While the empathetic agent module 56 and the LLM module 58 are provided with the infotainment system 22, the empathetic agent module 56 and / or the LLM module 58 may be provided in a separate device. For example, the empathetic agent module 56 and the LLM module 58 are disposed in the vehicle 10, but separate from the system 22. In another example, at least one of the empathetic agent module 56 and the LLM module 58 may be provided in cloud server and is communication with the vehicle 10.

[0063] Among other components, the infotainment system 22 of the present disclosure may include one or more processors and one or more non-transitory computer readable medium comprising programming instructions that when executed by the one or more processors, perform the functions described herein including those associated with the empathetic agent module 56, the LLM module 58, the speech interface module 36, the emotion detector module 32, the ASR module 38, and / or the TTS module 39.

[0064] While exemplary embodiments are described above, it is not intended that these embodiments describe all possible forms. Rather, the words used in the specification are words of description rather than limitation, and it is understood that various changes may be made without departing from the spirit and scope of the disclosure. Additionally, the features of various implementing embodiments may be combined to form further embodiments.

[0065] In this application, the term “module” (e.g., the empathetic agent module 56, the LLM module 58, the speech interface module 36, the emotion detector module 32, the ASR module 38, the TTS module 39, among other described herein modules) may refer to, be part of, or include:an Application Specific Integrated Circuit (ASIC); a digital, analog, or mixed analog / digital discrete circuit; a digital, analog, or mixed analog / digital integrated circuit; a combinational logic circuit; a field programmable gate array (FPGA); a processor circuit (shared, dedicated, or group) that executes code; a memory circuit (shared, dedicated, or group) that stores code executed by the processor circuit; other suitable hardware components that provide the described functionality; or a combination of some or all of the above, such as in a system-on- chip.

[0066] The term memory is a subset of the term computer-readable medium. The term computer- readable medium, as used herein, does not encompass transitory electrical or electromagnetic signals propagating through a medium (such as on a carrier wave); the term computer-readable medium may therefore be considered tangible and non-transitory. Non-limiting examples of a non- transitory, tangible computer-readable medium are nonvolatile memory circuits (such as a flash memory circuit, an erasable programmable read-only memory circuit, or a mask read only circuit), volatile memory circuits (such as a static random access memory circuit or a dynamic random access memory circuit), magnetic storage media (such as an analog or digital magnetic tape or a hard disk drive), and optical storage media (such as a CD, a DVD, or a Blu-ray Disc).

[0067] The systems, modules, and / or methods described in this application may be partially or fully implemented by a special purpose computer created by configuring a general-purpose computer to execute one or more particular functions embodied in computer programs. The functional blocks, flowchart components, and other elements described above serve as software specifications, which can be translated into the computer programs by the routine work of a skilled technician or programmer.

[0068] As used herein, the phrase at least one of A, B, and C should be construed to mean a logical (A OR B OR C), using a non-exclusive logical OR, and should not be construed to mean “at least one of A, at least one of B, and at least one of C.”

[0069] The description of the disclosure is merely exemplary in nature and, thus, variations that do not depart from the substance of the disclosure are intended to be within the scope of the disclosure. Such variations are not to be regarded as a departure from the spirit and scope of the disclosure

Claims

WHAT IS CLAIMED IS:

1. A system for regulating an emotional state of a driver of a vehicle, the system comprising: a processor; and a non-transitory computer readable medium comprising programming instructions that when executed by the processor, cause the processor to: operate a large language model; generate a prompt chain using an emotional state information providing the emotional state of the driver, the prompt chain including a sequence of parameterized prompts, wherein the sequence of parameterized prompts includes a first parameterized prompt that incorporates information gleaned from a response by the large language model to a second parameterized prompt that occurred prior to the first parameterized prompt; output, by the large language model, a candidate action to be undertaken to regulate the emotional state of the driver based the prompt chain and a context information; and cause the candidate action to be performed in response to receiving approval from the driver.

2. The system of claim 1, wherein the programming instructions further cause the processor to operate as an emotion detector that generates the emotional state information using uses information received from a cabin monitor provided in the vehicle.

3. The system of claim 1 , wherein the context information includes at least one of weather information, traffic information, or calendar information indicative of appointments for the driver.

4. The system of claim 1, wherein the context information includes driver preferences.

5. The system of claim 1, further comprising a cabin monitor configured to provide information used to generate the emotional state information, wherein the cabin monitor includes at least one of a camera, a microphone, or a physiological sensor.

6. The system of claim 1, wherein the programming instructions further cause the processor to form an aggregate of the emotional state information and the context information, wherein the aggregate is used by the large language model.

7. The system of claim 1, wherein the candidate action is selected to reduce exposure of a trigger to the driver.

8. The system of claim 1, wherein the candidate action includes defining a new travel route for the vehicle from a currently traveled route that exposes the driver to a trigger, wherein the trigger is anticipated to alter the emotional state of the driver from an optimal emotional state, and the new travel route lacks the trigger.

9. The system of claim 1, wherein the candidate action includes causing an infotainment system in the vehicle to play a sound that is anticipated to improve the emotional state of the driver, the sound is selected based on a stored preference of the driver.

10. The system of claim 1 , wherein the candidate action includes instructing the driver to carry out an exercise that is anticipated to improve the emotional state of the driver.

11. The system of claim 1, wherein the programming instructions further cause the processor to: observe the emotional state of the driver in response to a first candidate action being performed, wherein the first candidate action being the candidate action approved by the driver; recommend a new candidate action to the driver in response to the emotional state of the driver not changing or decreasing; and cause the new candidate action to be performed in response to receiving approval from the driver.

12. The system of claim 1, wherein the candidate action includes controlling a feature of the vehicle that is anticipated to improve the emotional state of the driver.

13. A method for regulating an emotional state of a driver of a vehicle, the method comprising: generating a prompt chain using an emotional state information providing the emotional state of the driver, the prompt chain including a sequence of parameterized prompts, wherein the sequence of parameterized prompts includes a first parameterized prompt that incorporates information gleaned from a response by a large language model to a second parameterized prompt that occurred prior to the first parameterized prompt; outputting, by the large language model, a candidate action to be undertaken to regulate the emotional state of the driver using the prompt chain and a context information; and causing the candidate action to be performed in response to receiving approval from the driver.

14. The method of claim 13, further comprising generating the emotional state information using information received from a cabin monitor provided in the vehicle, wherein the cabin monitor includes at least one of a camera, a microphone, or a physiological sensor.

15. The method of claim 13, wherein the context information includes at least one of weather information, traffic information, or calendar information indicative of appointments for the driver.

16. The method of claim 13, further comprising forming an aggregate of the emotional state information and the context information, wherein the aggregate is used by the large language model.

17. The method of claim 13, further comprising: observing the emotional state of the driver in response to a first candidate action being performed, wherein the first candidate action being the candidate action approved by the driver;recommending a new candidate action to the driver in response to the emotional state of the driver not changing or decreasing; and causing the new candidate action to be performed in response to receiving approval from the driver.

18. The method of claim 13, wherein the candidate action is selected to reduce exposure of a trigger to the driver.

19. The method of claim 13, wherein the candidate action includes defining a new travel route for the vehicle from a currently traveled route that exposes the driver to a trigger, wherein the trigger is anticipated to alter the emotional state of the driver from an optimal emotional state, and the new travel route lacks the trigger.

20. The method of claim 13, wherein the candidate action includes causing an infotainment system in the vehicle to play a sound that is anticipated to improve the emotional state of the driver, the sound is selected based on a stored preference of the driver.

Images