Devices and methods for controlling hardware agents in control scenarios with multiple hardware agents.
By combining neural networks and game theory layers, the problem of determining the common action sequence in multi-agent control was solved, and efficient and robust control was achieved in autonomous driving and multi-robot systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ROBERT BOSCH GMBH
- Filing Date
- 2021-08-13
- Publication Date
- 2026-05-26
AI Technical Summary
In the case of control by multiple hardware agents, existing technologies struggle to efficiently determine the common action sequence of each agent, take into account the goals and behaviors of other agents, and achieve robust control that maximizes benefits.
By using neural networks to determine the potential function and game theory layers, the common trajectory of multiple hardware agents is predicted. The first neural network determines the potential function parameters, the second neural network determines the control scenario, and the common action sequence is determined by searching for local optima. Control is then performed by combining sensor data and prior knowledge.
It achieves efficient and robust control of multiple hardware agents, can predict and optimize their common action sequence in real time, adapts to complex control situations, and is suitable for autonomous driving and multi-robot systems.
Smart Images

Figure CN114077242B_ABST