Motion planning for autonomous vehicles and reconfigurable motion planning processors

A motion planning device comprising:

• a hardware processor;
• memory storage;
• and
• a planning graph stored in the memory storage, the planning graph comprising a plurality of nodes connected by edges, wherein each node represents, implicitly or explicitly, time and variables defining a state of an autonomous vehicle, an operating environment, or both the state of the autonomous vehicle and the operating environment, wherein the hardware processor receives representations of obstacles that represent objects of an environment and sample trajectories of at least one of the obstacles that in the environment that is a dynamic object adjusts a respective representation of a probability of collision along each edge in the planning graph that results in a collision with the representation of the obstacles to account for the sample trajectories;
• and determines a path considering cost and probability of collision.

A method of performing motion planning for an autonomous vehicle comprising:

• receiving a representation of obstacles representing objects of an environment;
• sampling trajectories of any of the obstacles in the environment that represent dynamic objects of the environment;
• for each edge in a planning graph that results in a collision with the representations of obstacles, adjusting a respective cost value to account for a probability of collision based on the sample trajectories, wherein the planning graph comprises a plurality of nodes connected by edges, wherein each node represents, implicitly or explicitly, 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;
• and
• determining a path that considers cost and/or risk.

A method of operating a reconfigurable motion planning module, the method comprising:

• performing a first programming phase to configure a motion planning module for a particular autonomous vehicle and optionally a particular scenario, the first programming phase comprising:
• receiving a first planning graph;
• and for each edge of the first planning graph, receiving precomputed collision data for, and optionally cost associated with, that edge;
• storing the first planning graph and the precomputed collision data at a memory storage of the motion planning module;
• performing real-time motion planning using the motion planning module configured for the particular autonomous vehicle and optionally the particular scenario;
• and
• performing a second programming phase to configure the motion planning module for a new particular autonomous vehicle, a new particular scenario, or both a new particular autonomous vehicle and new particular scenario, the second programming phase comprising:
• receiving a second planning graph;
• and for each edge of the second planning graph, receiving precomputed collision data for, and optionally cost associated with, that edge;
• and
• storing the second planning graph and the precomputed collision data at the memory storage of the motion planning module.