Real time sensor based motion planning

Abstract

Robot motion planning opens areas of interests and challenges especially when attributed to sensor-based systems. Previous works mainly concern on dynamic and global optimal solution using what is proposed as on-line trajectory planner. The methods unfortunately often require intensive computation, and yet limit their practical implementation. Using simpler paradigm, i.e. kinematic optimality and optimal motion profile, it should be able to make the computation faster and generate optimal motion. For sensor based application, the motion profile is correlated to reactive motion planning. Typically trapezoidal profile fulfills kinematic optimality. The developed algorithms apply conditionals to select the corresponding motion parameter to equivalently generate trapezoidal motion profiles. The algorithm is found to be reliable, fast, and applicable in sensor based applications. The simulations and experiments show satisfactory results on typical system. However, implementation into a real higher multidimensional systems has not been realized.