Dynamic Goal Planning Using Sequence-to-Sequence Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing dynamic plans, especially those involving virtual assistants, lack the ability to adapt in real-time to changing circumstances, making them inefficient in complex, interdependent action sequences.
Innovation Solution
A system and method using machine learning, specifically a sequence-to-sequence model with a bidirectional recurrent neural network (RNN), to analyze and update goal plans by rearranging or modifying action sequences based on the current state, allowing for dynamic adaptation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a static action sequence is used in a goal plan, then the plan is simple to execute and finish, but the plan cannot adapt to changing circumstances or interdependent actions
Solution Approach 1:
The patent implements dynamic goal planning by enabling the system to continuously update and modify action sequences based on changing system states. The planner transitions from static pre-defined plans to dynamic adaptive plans that can be modified during execution, allowing virtual assistants to respond to new information and changing circumstances in real-time.
Solution Approach 2:
The system incorporates feedback mechanisms where the current state of the system is continuously monitored and fed back to the goal planner. This feedback loop enables the planner to assess whether actions are achieving their intended effects and to modify subsequent actions accordingly, creating an adaptive planning system that learns from execution outcomes.
2Productivity
If multiple interdependent actions are included in a plan, then the plan can accomplish complex goals, but the plan becomes difficult to optimize when states change
Solution Approach 1:
The system performs preliminary analysis of action interdependencies and potential state changes before executing the full action sequence. By pre-processing and understanding the relationships between actions, the system can quickly determine which actions need to be modified or skipped when states change, reducing the computational burden during real-time optimization.
Solution Approach 2:
The patent enables dynamic modification of action sequences by allowing the system to add, remove, or reorder actions based on current state assessments. This dynamic approach maintains productivity by keeping complex action sequences while simultaneously improving adaptability through real-time optimization capabilities.
3Ease of operation
If virtual assistants follow a fixed plan, then the execution is straightforward, but the assistants cannot change the plan based on real-time state changes
Solution Approach 1:
The system implements continuous feedback monitoring where the current state is assessed against the planned actions. When discrepancies are detected or new information becomes available, the system automatically updates the plan to maintain relevance, ensuring that virtual assistants execute reliable plans that reflect current conditions rather than outdated instructions.
Solution Approach 2:
The goal planner operates autonomously to update and modify action sequences without requiring external intervention. The system self-corrects and adapts plans based on state changes, maintaining ease of operation by automatically managing plan complexity while ensuring reliability through continuous state-based optimization.
Data Source
AI summary
The disclosed system and method provide a way to create, update, and execute dynamic goal plans. Updating a dynamic goal plan may be based on the initial sequence of actions of the goal plan as well as the corresponding states of the actions. By using a sequence to sequence model, a goal plan can still be processed when the length of the input (initial sequence of actions) differs from the length of the output (updated sequence of actions). A sequence to sequence model can determine the interdependencies between actions that can contribute to the optimal order in which actions can efficiently be performed. A single layer neural network or clustering can be used to approximate the state of a goal plan that may be capable infinite states. This approximation improves accuracy in capturing the state of a goal plan, thereby improving accuracy in predicting the future state of a system, which can help with planning (e.g., gathering resources in advance). Projects involving collaboration between virtual and/or human assistants can greatly benefit from the ability to update a dynamic goal plan in real time.