A collaborative cloud rendering system

Abstract

This invention discloses a collaborative cloud rendering system, including a lighting and shadow rendering module. The module's direct lighting rendering unit performs rendering through direct lighting calculations, the shadow rendering unit performs rendering through shadow calculations, the ambient occlusion rendering unit performs rendering through ambient occlusion calculations, and the indirect lighting rendering unit performs rendering through indirect lighting calculations. The indirect lighting rendering unit runs on a cloud backend device. A hybrid rendering unit performs rendering through hybrid calculations and runs on a web frontend device. Depending on the performance of the web frontend device, each unit runs on either the web frontend device or the cloud backend device. Therefore, this invention proposes four key real-time lighting and shadow rendering algorithms for collaborative real-time dynamic lighting and shadow rendering in Web3D cloud rendering, optimized through analysis of various data such as algorithm frame rate, the operating efficiency of the device running the algorithm, and the lighting and shadow rendering results.

Description

Technical Field

[0001] This invention relates to the field of cloud rendering technology, and in particular to a collaborative cloud rendering system. Background Technology

[0002] Moving applications from local machines to the network, running them on supercomputer clusters, and allowing user terminals to access the applications they need simply through a client—this is the concept of "cloud applications." Combining cloud applications with rendering has long been a research goal. As early as the 2009 International Consumer Electronics Show (CES), AMD pioneered the Fusion Render Cloud System (FRCS), a cloud rendering system that prioritizes cloud server rendering performance and optimizes real-time transmission of user commands. In 2013, researchers further discussed the construction methods of cloud rendering platforms and conducted performance evaluations. In recent years, with the popularization of online video games, cloud gaming, based primarily on cloud rendering technology, has become a major application area for current cloud rendering technology.

[0003] The existing technology has objective defects and the following technical problems:

[0004] Latency and stability issues: Because cloud rendering relies on network transmission, its latency and stability are often affected by network conditions, especially when rendering large scenes, which can impact rendering speed and quality. Cost issues: Cloud rendering requires the use of cloud computing resources, therefore it incurs corresponding costs. These costs can be very high, especially when rendering large scenes.

[0005] Security issues: Because cloud rendering requires uploading user scene files to cloud servers for rendering, data security is a concern. Without encryption or proper handling, user data can easily be stolen or suffer other security breaches.

[0006] Compatibility issues: Differences exist between different rendering engines and rendering software, which may cause cloud rendering services to be incompatible with all rendering software and rendering engines.

[0007] Controllability issues: Because cloud rendering requires uploading scene files to a cloud server, users cannot fully control the entire rendering process, nor can they monitor and adjust the rendering process in a timely manner, which may lead to some unpredictable problems. Summary of the Invention

[0008] To address the aforementioned shortcomings, the technical problem to be solved by this invention is to provide a collaborative cloud rendering system that proposes four key real-time lighting and shadow rendering algorithms for collaborative rendering of real-time dynamic lighting and shadow in Web3D cloud rendering. These algorithms are optimized through analysis of various data, including the algorithm's frame rate, the operating efficiency of the device on which the algorithm is located, and the lighting and shadow rendering results.

[0009] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0010] A collaborative cloud rendering system includes a web front-end device, a cloud back-end device, and a lighting and shadow rendering module. The lighting and shadow rendering module includes a direct lighting rendering unit, a shadow rendering unit, an ambient occlusion rendering unit, an indirect lighting rendering unit, and a hybrid rendering unit. The direct lighting rendering unit performs rendering through direct lighting calculation; the shadow rendering unit performs rendering through shadow calculation; the ambient occlusion rendering unit performs rendering through ambient occlusion calculation; the indirect lighting rendering unit performs rendering through indirect lighting calculation, and the indirect lighting rendering unit runs on the cloud back-end device; the hybrid rendering unit performs rendering through hybrid calculation, and the hybrid rendering unit runs on the web front-end device. Depending on the device performance of the web front-end device, the direct lighting rendering unit, the shadow rendering unit, and the ambient occlusion rendering unit run on the web front-end device or the cloud back-end device.

[0011] In a preferred embodiment, when the Web front-end device is a low-performance device, the hybrid rendering unit and a portion of the direct lighting rendering unit run on the Web front-end device, while the other portion of the direct lighting rendering unit, the shadow rendering unit, the ambient occlusion rendering unit, and the indirect lighting rendering unit all run on the cloud back-end device.

[0012] The preferred method is that the irradiance contained in the pixels of the final output rendering frame of the cloud rendering system is: Where I′ DL W represents the local illumination rendering radiance of a web front-end device. front The weight of the front-end illumination radiation. V represents the local illumination rendering radiance of cloud backend devices. SH and V AO I refers to the visibility of the current light after shadow rendering and ambient occlusion rendering. IL This is the relevant irradiance information obtained after indirect lighting rendering.

[0013] In a preferred embodiment, when the Web front-end device is a high-performance device, the direct lighting rendering unit, the shadow rendering unit, the ambient occlusion rendering unit, and the hybrid rendering unit all run on the Web front-end device; the indirect lighting rendering unit runs on the cloud back-end device.

[0014] The preferred embodiment is that the irradiance contained in the pixels of the final output rendering frame of the cloud rendering system is: I = I DL V SH V SO W front +I IL W back , where I DL It is the related irradiance information obtained after rendering the direct illumination irradiance, V SH and V AO I refers to the visibility of the current light after shadow rendering and ambient occlusion rendering. IL W represents the relevant irradiance information obtained after indirect lighting rendering. front W is the weight of the front-end illuminance. back It is the weight obtained from the back-end irradiance.

[0015] The preferred method is that the direct lighting algorithm includes an ambient lighting algorithm and a traditional lighting algorithm; the direct lighting algorithm is set as an ambient lighting algorithm on low-performance web front-end devices, as a traditional lighting algorithm on high-performance web front-end devices, and as a traditional lighting algorithm on the cloud back-end devices.

[0016] The preferred embodiment is that the shadow algorithm includes a shadow map algorithm and a variance shadow map algorithm; the shadow algorithm is not set on low-performance web front-end devices, is set as the shadow map algorithm on high-performance web front-end devices, and is set as the variance shadow map algorithm on the cloud back-end devices.

[0017] The preferred embodiment is that the ambient occlusion algorithm includes the Screen space ambient occlusion algorithm and the Voxel accelerate ambient occlusion algorithm; the ambient occlusion algorithm is not set on low-performance Web front-end devices, the Screen space ambient occlusion algorithm is set on high-performance Web front-end devices, and the Voxel accelerate ambient occlusion algorithm is set on cloud back-end devices.

[0018] The preferred method is that the indirect lighting algorithm is the Voxel cone tracing algorithm; the indirect lighting algorithm is set to the Voxel cone tracing algorithm on the cloud backend device.

[0019] After adopting the above technical solution, the beneficial effects of the present invention are:

[0020] The collaborative cloud rendering system of this invention includes a web front-end device, a cloud back-end device, and a lighting and shadow rendering module. The lighting and shadow rendering module includes a direct lighting rendering unit, a shadow rendering unit, an ambient occlusion rendering unit, an indirect lighting rendering unit, and a hybrid rendering unit. The direct lighting rendering unit performs rendering through direct lighting calculations; the shadow rendering unit performs rendering through shadow calculations; the ambient occlusion rendering unit performs rendering through ambient occlusion calculations; the indirect lighting rendering unit performs rendering through indirect lighting calculations and runs on the cloud back-end device; the hybrid rendering unit performs rendering through hybrid calculations and runs on the web front-end device. Depending on the device performance of the web front-end device, the direct lighting rendering unit, shadow rendering unit, and ambient occlusion rendering unit run on either the web front-end device or the cloud back-end device. Therefore, this invention proposes four key real-time lighting and shadow rendering algorithms for collaborative real-time dynamic lighting and shadow rendering in Web3D cloud rendering. These algorithms are ultimately optimized through analysis of various data, including the algorithm's frame rate, the operating efficiency of the device running the algorithm, and the lighting and shadow rendering results. Detailed Implementation

[0021] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions in the embodiments of this invention are described clearly and completely below. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0022] A collaborative cloud rendering system includes a web front-end device, a cloud back-end device, and a lighting and shadow rendering module. The lighting and shadow rendering module includes a direct lighting rendering unit, a shadow rendering unit, an ambient occlusion rendering unit, an indirect lighting rendering unit, and a hybrid rendering unit. The direct lighting rendering unit performs rendering through direct lighting calculation; the shadow rendering unit performs rendering through shadow calculation; the ambient occlusion rendering unit performs rendering through ambient occlusion calculation; the indirect lighting rendering unit performs rendering through indirect lighting calculation, and the indirect lighting rendering unit runs on the cloud back-end device; the hybrid rendering unit performs rendering through hybrid calculation, and the hybrid rendering unit runs on the web front-end device. Depending on the device performance of the web front-end device, the direct lighting rendering unit, shadow rendering unit, and ambient occlusion rendering unit run on either the web front-end device or the cloud back-end device.

[0023] It is important to note that a device's rendering capability is related to its hardware performance; the stronger the hardware, the stronger the rendering capability, and vice versa. Just as a desktop PC with powerful rendering capabilities is often equipped with a high-performance graphics card, which has higher power consumption and stricter heat dissipation requirements, mobile devices, constrained by heat dissipation, battery (limiting overall performance), and physical space, are obviously far inferior to desktop PCs in performance and rendering capabilities. Therefore, this system also categorizes the aforementioned web front-end hardware devices into two types based on their performance: low-performance mobile devices primarily powered by batteries (mobile phones, tablets, VR headsets (with rendering kernels), and low-performance mobile PCs) and high-performance devices primarily powered by external power supplies (high-performance mobile PCs and high-performance desktop PCs).

[0024] In some embodiments of the present invention, when the Web front-end device is a low-performance device, the hybrid rendering unit and a portion of the direct lighting rendering unit run on the Web front-end device, while the other portion of the direct lighting rendering unit, shadow rendering unit, ambient occlusion rendering unit, and indirect lighting rendering unit all run on the cloud back-end device.

[0025] When a web front-end runs on low-performance hardware, its hardware rendering capabilities are limited. Therefore, in the cloud baking system, the basic lighting and rendering tasks running on the web front-end are set to direct lighting rendering, while most of the computationally intensive rendering tasks, such as shadow rendering, ambient occlusion rendering, and indirect lighting rendering, are executed on the cloud back-end. However, since these rendering tasks on the cloud back-end all rely on local lighting information, some direct lighting rendering is also run on the cloud back-end.

[0026] The aforementioned mechanism allows the cloud backend to encompass almost all important rendering tasks, essentially making it equivalent to the tasks performed by a remote rendering system on the server side. However, even under this setup, the cloud baking system still possesses two advantages that remote rendering systems cannot achieve:

[0027] (1) Even if the front end does not receive the light map from the back end in time, the Web front end still has local lighting rendering effects displayed, so that rendering distortion such as "frame freeze" and "black screen" caused by missing rendering frames will not occur; (2) The cloud back end can reduce the rendering frame rate of the cloud back end by means of the "rendering latency optimization strategy", which greatly reduces the resource consumption of the rendering server in the cloud.

[0028] In summary, the formula for calculating the final illuminance I obtained from lighting and shadow rendering is as follows:

[0029] I = I front W front +I back W back .

[0030] Among them I front It is the illumination radiation obtained from web front-end lighting and shadow rendering, W front It is the weight of the front-end illuminance, I back It refers to the illumination radiation of the cloud server backend rendering, W. back This is the weight derived from the back-end irradiance. The "I" in the following text... DL and I IL These are the irradiance information obtained after direct lighting rendering and indirect lighting rendering, respectively, and their values ​​are stored in the direct lighting map and the indirect lighting map. V SH , V AO This refers to the visibility of the current light after shadow rendering and ambient occlusion rendering. Its value is between 0.0 and 1.0, and it is stored in the shadow value map and ambient light map, respectively.

[0031] Based on the lighting and shadow scheduling settings of low-power web front-end devices, the irradiance I obtained from web front-end lighting and shadow rendering can be determined. front The calculation formula is:

[0032] I front =I′ DL .

[0033] I' DL This represents the local illumination rendering radiance of the web front-end, while I^ DL This represents the local illumination rendering radiance of the cloud backend, both of which differ from the traditional local illumination rendering radiance I. DL The differences are as follows, and the irradiance I rendered by the cloud server backend is... back The calculation formula is:

[0034]

[0035] Therefore, it can be deduced that the irradiance contained in the pixels of the final output rendered frame of this mechanism is:

[0036]

[0037] It can be seen from the above formula that:

[0038] (1) When the Web front-end of the rendering system uses a low-performance device, the direct lighting settings of the front-end and back-end of the system are different; (2) The pixels in the shadow map or ambient occlusion map will be multiplied by the pixels of the direct lighting map to obtain the corresponding effect; (3) The indirect lighting rendering map only needs to add the pixel value directly to the pixel value of other rendering effect maps; (4) The final output rendering frame is obtained by directly adding the pixel values ​​of the Web rendering image frame and the cloud light map.

[0039] In other embodiments of the present invention, when the Web front-end device is a high-performance device, the Web front-end will undertake more rendering tasks, and the direct lighting rendering unit, shadow rendering unit, ambient occlusion rendering unit and hybrid rendering unit will all run on the Web front-end device; the indirect lighting rendering unit will run on the cloud back-end device.

[0040] The reasons for this setting include: (1) With the assistance of Three.js, high-performance PCs can render hard shadows and ambient occlusion effects in real time (frame per second (FPS) greater than 30); (2) Indirect lighting rendering, as a complex lighting effect with high computational requirements, will still be placed on the cloud backend.

[0041] Therefore, based on the lighting and shadow scheduling settings of high-energy-consuming web front-end devices, the irradiance I obtained from web front-end lighting and shadow rendering can be determined. front The calculation formula is:

[0042] I front =I DL V SH V SO .

[0043] The formula for calculating the irradiance of the cloud server backend rendering is:

[0044] I back =I IL .

[0045] Therefore, it can be deduced that the irradiance contained in the pixels of the final output rendered frame of this mechanism is:

[0046] I = I DL V SH VSO W front +I IL W back , where I DL It is the related irradiance information obtained after rendering the direct illumination irradiance, V SH and V AO I refers to the visibility of the current light after shadow rendering and ambient occlusion rendering. IL W represents the relevant irradiance information obtained after indirect lighting rendering. front W is the weight of the front-end illuminance. back It is the weight obtained from the back-end irradiance.

[0047] It can be seen from the above formula that:

[0048] (1) The front end has completed the rendering of local lighting and related shadow effects; (2) The back end only includes the rendering of indirect lighting. Compared with the cloud baking system where the front end is a low-power device, the rendering task of the cloud back end is greatly reduced.

[0049] After determining the device performance-oriented lighting and shadow rendering mechanism, this system selected and tested suitable lighting and shadow rendering algorithms for low-performance web front-end devices, high-performance web front-end devices, and cloud back-end devices (device configurations are shown in Table A1). Considering the frame rate requirements of Web3D applications, the basic rendering frame rate for 3D scene lighting and shadow rendering was set to no less than 30 FPS; the test scene was set as a 3D scene—Sponza; the test parameters included the frame rate, CPU / GPU utilization, and memory utilization during full-frame rendering; the tested lighting and shadow algorithms were divided into four categories: direct lighting algorithm, shadow algorithm, ambient occlusion algorithm, and indirect lighting algorithm.

[0050] In this system, two direct lighting algorithms are configured: one is the Ambient lighting algorithm, which has extremely low rendering computational requirements, and the other is the traditional Blinnphong (BP) lighting algorithm, which includes ambient lighting, diffuse lighting, and specular lighting. Table 1 shows that when the Ambient algorithm is deployed on a low-performance web front-end device, the frame rate reaches 46 FPS, and the CPU / GPU and memory usage are within reasonable ranges (CPU usage not exceeding 50%, GPU usage not exceeding 90%, and memory usage not exceeding 2GB are considered reasonable). Subsequently, the BP algorithm was tested on low-performance web front-end devices, high-performance web front-end devices, and cloud back-end devices. It was found that the algorithm could not run smoothly on the low-performance web front-end device, with a frame rate lower than the basic frame rate setting for the front-end, while running smoothly on the other two sets of devices. Therefore, the direct lighting in this system is configured with the Ambient algorithm on low-performance web front-end devices, the BP algorithm on high-performance web front-end devices, and the BP algorithm on cloud back-end devices, as shown in Table 2.

[0051] Table 1 Performance of lighting and shadow rendering algorithms on different devices

[0052] Table 1 The performance of lighting and shadow rendering algorithms in different devices

[0053]

[0054] Table 2. Lighting and shadow rendering algorithm adaptation based on device performance.

[0055] Table 2 The lighting and shadow algorithm selection based on device configuration

[0056]

[0057] This system employs two real-time shadow algorithms: Shadow Map (SM), which consumes less rendering power and primarily uses hard shadows, and Variance Shadow Map (VSM), which consumes more rendering power and primarily uses soft shadows. As shown in Table 1, deploying the SM algorithm on a low-performance web front-end device resulted in a frame rate of less than 18 FPS, indicating that deploying the SM algorithm on such devices is unreasonable. Subsequently, testing the SM algorithm on a high-performance web front-end device revealed smooth operation with a frame rate of 65 FPS, and hardware resource consumption remained within a reasonable range. Finally, testing the more complex and realistic soft shadow algorithm VSM on both high-performance web front-end devices and cloud back-end devices revealed that VSM could only achieve 22 FPS on the high-performance front-end device, failing to run smoothly; only on the cloud back-end could it achieve a high frame rate. Therefore, the shadow algorithm of this system is not set on low-performance Web front-end devices, the SM algorithm is set on high-performance Web front-end devices, and the VSM algorithm is set on cloud back-end devices, as shown in Table 2.

[0058] Ambient occlusion, a part of global illumination rendering algorithms, reflects the occlusion details of indirect lighting on the model, resulting in excellent realism. This system uses two ambient occlusion algorithms: Screen Space Ambient Occlusion (SSAO), which has lower computational complexity and is calculated based on screen space rendering targets; and Voxel Accelerate Ambient Occlusion (VAAO), which has higher computational complexity but stronger realism and is based on sparse space voxels. As shown in Table 1, deploying the SSAO algorithm on low-performance web front-end devices resulted in poor performance, with a frame rate of only 23 FPS, below the basic rendering frame rate requirement. High-performance web front-end devices, however, supported it well. The more realistic VAAO algorithm has higher computational complexity than SSAO, so it was only deployed on high-performance web front-end devices and cloud back-end devices. Testing revealed that high-performance web front-end devices running this algorithm achieved a frame rate of less than 20 FPS, and hardware resource consumption significantly exceeded reasonable limits, while the cloud back-end ran the algorithm smoothly at high frame rates. Therefore, the ambient light occlusion algorithm of this system is not set on low-performance Web front-end devices, the SSAO algorithm is set on high-performance Web front-end devices, and the VAAO algorithm is set on cloud back-end devices, as shown in Table 2.

[0059] Indirect lighting is achieved by using the diffuse and specular colors generated under direct lighting as secondary light sources to reflect off other object surfaces. Due to its high algorithmic complexity and computational demands, indirect lighting is only deployed on the cloud system in this system. In recent years, various methods for real-time rendering of indirect lighting have emerged, with prominent ones including Reflective Shadow Map (RSM), Light Propagation Volume (LPV), and Voxel Contracing (VCT). Among these, VCT is widely used due to its superior rendering results and efficiency, and this system employs this algorithm. The test results in Table 1 show that the cloud backend can run this algorithm smoothly at high frame rates. Therefore, the indirect lighting algorithm in this system is only set to VCT on the cloud backend device, as shown in Table 2.

[0060] In summary, this test, using the classic Sponza model as the test object, fully elucidates the basic ideas and implementation process of the system's lighting and shadow rendering through the adaptation of four types of rendering algorithms with three major devices in the system. The system starts with direct lighting, enabling the scene to have local illumination first; then, the system adds the calculation of lighting visibility, which includes two parts: the shadow effect caused by direct lighting occlusion and the ambient light occlusion effect caused by indirect lighting occlusion; finally, the system implements the most complex indirect lighting algorithm. As can be seen from Table 2, for low-performance web front-end devices, only the Ambient direct lighting algorithm with extremely low algorithm complexity needs to be implemented, and other rendering effects need to be assisted by cloud rendering; while for high-performance web front-end devices, three types of lighting and shadow rendering, including the Blinn phong direct lighting algorithm, the Shadow map algorithm, and the Screen space ambient occlusion algorithm, are implemented, and cloud rendering only needs to assist in the implementation of its indirect lighting.

[0061] The advantage of this invention is that in this system, all pixels of the image at viewpoint p are transformed into the image captured at viewpoint p′ through matrix transformation. This rendered image is not generated from a backend rendering, but rather through a secondary projection transformation of the received cloud rendering image. For areas visible in the new viewpoint p′ but invisible in the reference viewpoint p, the newly generated image automatically fills them with black pixels, resulting in a "black hole" distortion phenomenon. This system employs a texture sampling-based hole-filling algorithm to address this issue.

[0062] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications or improvements to an equivalent collaborative cloud rendering system made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A collaborative cloud rendering system, characterized in that, It includes a web front-end device, a cloud back-end device, and a lighting and shadow rendering module. The lighting and shadow rendering module includes a direct lighting rendering unit, a shadow rendering unit, an ambient light occlusion rendering unit, an indirect lighting rendering unit, and a hybrid rendering unit. The direct lighting rendering unit performs rendering through direct lighting calculations; The shadow rendering unit renders shadows through shadow calculation; The ambient occlusion rendering unit performs rendering based on ambient occlusion calculations. The indirect lighting rendering unit performs rendering through indirect lighting calculation, and the indirect lighting rendering unit runs on the cloud backend device; The hybrid rendering unit performs rendering through hybrid computation, and the hybrid rendering unit runs on the web front-end device; When the Web front-end device is a low-performance device, the hybrid rendering unit and a portion of the direct lighting rendering unit run on the Web front-end device, while the other portion of the direct lighting rendering unit, the shadow rendering unit, the ambient occlusion rendering unit, and the indirect lighting rendering unit all run on the cloud back-end device. When the Web front-end device is a high-performance device, the direct lighting rendering unit, the shadow rendering unit, the ambient occlusion rendering unit, and the hybrid rendering unit all run on the Web front-end device; the indirect lighting rendering unit runs on the cloud back-end device.

2. The collaborative cloud rendering system according to claim 1, characterized in that, The irradiance contained in the pixels of the final rendered frame output by the cloud rendering system is: Where I′ DL W represents the local illumination rendering radiance of a web front-end device. front The weight of the front-end illumination radiation. V represents the local illumination rendering radiance of cloud backend devices. SH and V AO I refers to the visibility of the current light after shadow rendering and ambient occlusion rendering. IL This is the relevant irradiance information obtained after indirect lighting rendering.

3. The collaborative cloud rendering system according to claim 1, characterized in that, The irradiance contained in the pixels of the final rendered frame output by the cloud rendering system is: I = I DL V SH V SO W front +I IL W back , where I DL It is the related irradiance information obtained after rendering the direct illumination irradiance, V SH and V AO I refers to the visibility of the current light after shadow rendering and ambient occlusion rendering. IL W represents the relevant irradiance information obtained after indirect lighting rendering. front W is the weight of the front-end illuminance. back It is the weight obtained from the back-end irradiance.

4. The collaborative cloud rendering system according to any one of claims 1 to 3, characterized in that, Direct lighting algorithms include ambient lighting algorithms and traditional lighting algorithms; The direct lighting algorithm is set as the ambient lighting algorithm on low-performance web front-end devices, as the conventional lighting algorithm on high-performance web front-end devices, and as the conventional lighting algorithm on the cloud back-end devices.

5. The collaborative cloud rendering system according to any one of claims 1 to 3, characterized in that, Shadow algorithms include shadow map algorithm and variance shadow map algorithm; The shadow algorithm is not configured on low-performance web front-end devices, but is configured as the Shadowmap algorithm on high-performance web front-end devices, and as the Variation shadow map algorithm on cloud back-end devices.

6. The collaborative cloud rendering system according to any one of claims 1 to 3, characterized in that, Ambient occlusion algorithms include the Screen space ambient occlusion algorithm and the Voxel accelerate ambient occlusion algorithm; The ambient occlusion algorithm is not configured on low-performance web front-end devices, but is configured as the Screen space ambient occlusion algorithm on high-performance web front-end devices, and as the Voxel accelerateambient occlusion algorithm on cloud back-end devices.

7. The collaborative cloud rendering system according to any one of claims 1 to 3, characterized in that, The indirect lighting algorithm uses the Voxel cone tracing algorithm; the indirect lighting algorithm is set to the Voxel cone tracing algorithm on the cloud backend device.

Images