Reconfigurable Motion Planning Processors for Real-Time Collision Checks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in performing real-time motion planning in environments with obstacles, particularly in handling dynamic obstacles and replanning at high frequencies, due to the computational intensity of collision checks.
Innovation Solution
A reconfigurable processor-based motion planning system that performs collision detection and path planning entirely in hardware or software, or a combination of both, using a motion planning module with a planning graph that adjusts collision probabilities based on perception data and trajectory samples, enabling risk-aware motion planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion planning algorithms perform comprehensive collision checks for real-time path planning, then collision avoidance reliability is improved, but computational time and processing power requirements increase
Solution Approach 1:
The patent segments the motion planning problem into multiple independent collision check operations that can be executed in parallel. The planning graph is divided into multiple edges that can be evaluated simultaneously by parallel processing units, allowing comprehensive collision detection without sequential processing delays.
Solution Approach 2:
The patent performs preliminary collision detection by pre-calculating collision probabilities for graph edges before final path selection. By evaluating collision risks in advance and storing results, the system avoids redundant computation during real-time path planning while maintaining reliable collision avoidance.
2Reliability
If motion planning algorithms increase collision check frequency for dynamic obstacles, then safety in dynamic environments is improved, but computational load increases
Solution Approach 1:
The patent implements periodic collision detection by evaluating collision probabilities at regular intervals along the planning graph edges. This periodic approach ensures dynamic obstacles are monitored continuously for safety while allowing computational resources to be managed efficiently through structured, interval-based processing.
Solution Approach 2:
The patent uses probabilistic roadmaps that can be copied and reused for multiple planning queries. Once a planning graph is constructed with collision probability data, it can be replicated and used for various path planning tasks, reducing the need to perform complete collision analysis from scratch for each new query.
3Adaptability or versatility
If a reconfigurable processor is used for motion planning, then adaptability to different robots and scenarios is improved, but device complexity increases
Solution Approach 1:
The patent employs a reconfigurable processor that can dynamically change its hardware configuration based on the specific robot and scenario requirements. The processing architecture can be reprogrammed to match the kinematic model, obstacle characteristics, and planning parameters of different applications, providing adaptability while managing complexity through systematic reconfiguration approaches.
Solution Approach 2:
The patent designs a universal motion planning processor that can handle multiple robot types and scenario variations through a single reconfigurable architecture. By creating a multi-functional platform that can be adapted to different applications, the system achieves versatility without requiring separate dedicated hardware for each robot type or scenario.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6D
AI summary
A system for motion planning for autonomous vehicles can include a plurality of sensors, a plurality of detectors in electrical communication with the plurality of sensors, and a motion planning module in electrical communication with the plurality of detectors and a computing system of an autonomous vehicle. The motion planning module stores a planning graph with each node representing, explicitly or implicitly, time and variables defining a state of the autonomous vehicle, an operating environment, or both the state of the autonomous vehicle and the operating environment. A reconfigurable processor can include a collision detection module and, optionally, a shortest path module. Pre-computed collision data and planning graph data reflecting logical/physical node mapping can be communicated to the processor during a programming phase and used during runtime.