Real-Time Dynamic Trajectory Optimization

This work revolves around planning fast, efficient intercept trajectories that enable a free-flying space robot to capture a maneuvering target vehicle in an obstacle filled environment.

One of our free flying robots (Huey) about to catch an object moving across the table.

Thesis Abstract

The capability of robots to complete tasks or entire missions autonomously relies heavily on their ability to plan. Good planners must not only be able to produce efficient plans but must also be able to quickly modify those plans in response to unpredicted events. Unfortunately these two goals are often at odds, with only slow, complex planners able to produce efficient plans, and only quick, simple planners able to react to unpredicted events.

This presentation examines the field of trajectory planning, for which this problem is especially prevalent. Trajectory planning is the process of generating dynamically consistent plans for autonomous robot locomotion. Trajectory planners can be categorized as either off-line planners which produce highly efficient trajectories but operate slowly in a plan-then-execute fashion, or as reactive planners which operate quickly in a plan-while-executing fashion but are unable to produce highly efficient trajectories.

The subject of this presentation is the development of a real-time trajectory optimization system that provides both highly efficient trajectories and the capability to react to uncertainty in the environment. This novel real-time trajectory optimization system achieves these capabilities by utilizing simultaneous planning and execution to improve the robot's trajectory while the robot is moving along it. Improved trajectory information is sent from the planner to the robot and data about the environment is sent from the robot to the planner in real-time. The system synchronizes the flow of this data with its planning process so that the operation of the system is fast and efficient, even as unpredicted events take place.

It is also shown that when considering the total system performance, which includes the planning duration as well as the trajectory duration, the real-time trajectory optimization system provides significantly better performance than off-line trajectory optimization systems, the previous standard for comparison. This performance increase is analyzed as a function of relative planning and execution speed and is shown to be up to 50% for a range of problems.

Finally, experimental results show the real-time trajectory optimization system controlling the motion of a thruster propelled free-flying robot. The system allows the robot to efficiently maneuver around obstacles and intercept an unpredictably maneuvering target vehicle.

Here are some MPEG Videos of Huey maneuvering.

An example set of trajectories generated while the robot was trying to chase down a moving target while avoiding an obstacle in it's path.

Back to the ARL Home Page