Autonomous Vehicle Motion Planning With Dynamic Collision Probabilities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in motion planning, particularly in environments with dynamic objects, where predicting the movements of these objects is uncertain, making it difficult to avoid collisions or intentionally interact with them in real-time.
Innovation Solution
A processor-based system that uses a reconfigurable collision detection architecture to assess collisions and modify edge weights in a planning graph based on the probability of collisions with dynamic objects, allowing for optimization of paths that either minimize or maximize collision potential, depending on the vehicle's goals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the autonomous vehicle performs real-time motion planning to avoid collisions with dynamic objects, then collision avoidance capability is improved, but the system complexity and computational burden increase
Solution Approach 1:
The motion planning system is segmented into distinct functional modules: perception module for detecting dynamic objects, prediction module for forecasting object trajectories, planning module for generating collision-free paths, and control module for executing motion commands. This modular segmentation allows each component to specialize in specific tasks, improving overall collision avoidance reliability while managing system complexity through clear separation of concerns
Solution Approach 2:
The system performs preliminary actions by pre-processing perception data to identify dynamic objects and their characteristics before formal motion planning begins. The prediction module pre-calculates probable trajectories of dynamic objects based on current motion states, allowing the planning module to work with pre-prepared trajectory information rather than raw sensor data, thus reducing real-time computational burden
2Measurement precision
If the autonomous vehicle uses complex collision detection algorithms to account for uncertain movements of dynamic objects, then collision detection accuracy is improved, but computational time and processing power requirements increase
Solution Approach 1:
The system dynamically changes parameters based on environmental context and object characteristics. The prediction horizon is adjusted according to object speed and distance - faster or closer objects receive longer prediction horizons for higher accuracy, while slower or farther objects use shorter horizons to reduce computation. Similarly, the resolution of collision detection grids and the number of simulated trajectories are adaptively adjusted based on risk levels, maintaining high accuracy for critical situations while reducing computational load in low-risk scenarios
Solution Approach 2:
The system applies partial collision detection by focusing computational resources on critical areas rather than performing exhaustive detection everywhere. The planning space is divided into regions of interest based on dynamic object positions and predicted trajectories, with higher detection accuracy applied only to regions where collisions are probable. This selective application of complex algorithms maintains collision detection accuracy for critical paths while reducing overall computational time
3Productivity
If the autonomous vehicle optimizes motion plans to account for fuel usage and other costs, then operational efficiency is improved, but the complexity of path optimization increases
Solution Approach 1:
The cost function parameters are dynamically adjusted based on operational context. Fuel consumption weights are increased during long-distance travel or when battery charge is low, while maneuvering costs become more significant in urban environments with frequent stops. This adaptive parameter adjustment allows the optimization algorithm to focus computational effort on the most relevant cost factors for each situation, improving operational efficiency without requiring the system to constantly evaluate all possible cost parameters at full complexity
Solution Approach 2:
The optimization problem is made dynamic by continuously updating cost parameters and constraints based on changing environmental conditions and vehicle state. As the vehicle moves through different environments (open road, urban area, parking lot), the relative importance of different cost factors changes dynamically. The optimization algorithm adapts to these changing conditions by adjusting its search strategy and parameter weights in real-time, improving operational efficiency while managing complexity through dynamic rather than static optimization
4Adaptability or versatility
If the autonomous vehicle performs real-time replanning to keep up with changes in the environment, then adaptability to dynamic changes is improved, but computational load and processing time increase
Solution Approach 1:
The motion planning system performs periodic replanning at fixed time intervals or after traveling fixed distances, rather than continuously replanning in response to every environmental change. Between periodic replanning events, the system executes the current plan while monitoring for significant deviations. This periodic approach maintains adaptability to dynamic changes by regularly updating the plan based on current conditions while avoiding the excessive computational load of continuous replanning
Solution Approach 2:
The system performs preliminary validation of potential replanning triggers before initiating full replanning sequences. When environmental changes are detected, the prediction module first evaluates whether the change is significant enough to warrant replanning by comparing predicted future states against safety thresholds. Only when changes exceed predefined significance levels does the system initiate computationally intensive replanning, thus maintaining adaptability to important changes while reducing unnecessary replanning computations
Data Source
AI summary
A motion planner of an autonomous vehicle's computer system uses reconfigurable collision detection architecture hardware to perform a collision assessment on a planning graph for the vehicle prior to execution of a motion plan. For edges on the planning graph, which represent transitions in states of the vehicle, the system sets a probability of collision with a dynamic object in the environment based at least in part on the collision assessment. Depending on whether the goal of the vehicle is to avoid or collide with a particular dynamic object in the environment, the system then performs an optimization to identify a path in the resulting planning graph with either a relatively high or relatively low potential of a collision with the particular dynamic object. The system then causes the actuator system of the vehicle to implement a motion plan with the applicable identified path based at least in part on the optimization.