Local Route Replanning for Agricultural Robots


In this project, we design algorithms to allow for efficient replanning for route-planners by bending trajectories around obstacles.


We present a new scheme for an offline route planner for coverage of an agricultural area, which is capable of local modifications in reaction to obstacles. For a given vehicle dynamics and specifications, the route planner generates optimal routes which not only maximise the area covered in a given time but also respect all constraints. The paper uses the help of spline functions to model curved trajectories as well as provide a bound on the curvature of the route. The issue of turning at headlands is also taken up and a framework is created to enable turning at any angle avoiding "no go" areas. Finally the paper considers newer tactics of cellular traversal, providing guaranteed better results in bounded time than current methods.


  • S. Choudhury "A Coverage Planning Algorithm for Agricultural Robots", Technical Report, IIT Kharagpur (pdf).

  • Details

    Here is a talk I gave at CMU and IIT on this work. The highlights of the description are provided below.

    The foundation of the approach is built on the Split and Merge method to route plan for non-convex fields. An animation of the method is provided below

    When obstacles appear on this field, the routes have to be replanned. While traditional methods have to completely replan (as shown on left) , our method bends the tracks (right).

    This bending is done using a gradient optimizer over a potential field. A spline is used for smooth optimization.

    Another way explored was partially following field boundary and partially covering field.