Vision-based Motion Planning for an Autonomous Motorcycle on Ill-Structured Road Reporter ：鄒嘉恆 Date ： 08/31/09

Intr oduction Design a vision-based motion planning system for desert terrain. The motion planning is based on a vision vector space(V 2 -space). Motivated by the DARPA Grand Challenge.

Outline V 2 -space Algorithm – V 2 -Space construction – Motion planning in V 2 -space Experimental results Conclusion

V2-space Pin-hole model ： Video frame F is a matrix of RGB values ：

Algorithm ： V2-space construction

Color correction ：

Algorithm ： V2-space construction Surface verification ： Obstacles detection ：

Algorithm ： V2-space construction Direction extraction ：

Algorithm ： Motion planning in V2-space A point robot with no time delay ：

Algorithm ： Motion planning in V2-space Choose the velocity profile v p (t) The motorcycle cannot run too fast for a given trajectory radius R.

A lgorithm ： Motion planning in V 2 -space Incorporating GPS information ：

Algorithm ： Vehicle size and processing delay Vehicle size

Algorithm ： Vehicle size and processing delay Image processing delay ：

Algorithm overview

Experimental results From video clip ：

Experimental results From video clip ：

Experimental results From video clip ：

Experimental results Filed tests ：

Conclusions Propose V 2 -space, a new framework that represent road feature and allows fast construction and motion planning. In the future, they will consider incorporating V 2 -space in a stereo vision system.