IEEE ICRA 2012 Needle Steering Workshop

Medical Needle Steering under Motion and Sensor Noise using Feedback-based Information Roadmaps

Ali-akbar Agha-mohammadi, Suman Chakravorty, Nancy M. Amato

Texas A&M University 

Abstract

Robot-assisted motion planning for medical needle steering under motion uncertainty and under noisy sensor measurements is an instance of the sequential decision making problem in partially-observable environments, which can be modeled as a Partially-Observable Markov Decision Process (POMDP) problem in its most general form. In this work, we propose using Feedback-based Information RoadMap (FIRM) to address the needle steering problem in continuous state, control, and observation spaces. FIRM-based needle steering offers more robustness and safety in the steering procedure. The FIRM-based planner is more robust since the solution of the planning algorithm is a feedback over the entire workspace of the needle, in contrast to the case that the solution of planning is a fixed nominal path, which has to be tracked by a controller. This property is rooted in the fact that FIRM induces reachable nodes in belief space, which results in a query-independent roadmap in belief space with independent edge costs and independent local planners. Moreover, FIRM-based needle steering offers more reliable plans and safety as it is able to compute more accurate collision probabilities that do not require typical simplifying assumptions such as the collision events in different time-instances are independent. We demonstrate different instantiations of the FIRM method using different controllers and show the planning performance on a planar needle model, and compare the robustness and safety obtained by this method to existing methods. Also, we discuss the existing challenges and future directions.

A simple environment is shown, where obstacles are depicted by gray shapes. A simple sensor model is also considered for illustration purposes, in which the sensors provide noisy range and bearing measurements from the landmarks, shown by black stars. The standard deviation of the distribution of the sensor noise increases linearly with the distance from landmarks. Left figure shows the sampled Gaussian belief nodes; the local planners are designed such that these belief nodes can be reached independent of the initial belief of the local planner. Middle figure shows the uncertainty propagation under the local planners; the crucial advantage of the method with respect to the state-of-art is that the uncertainty propagation on each edge is independent of the path that has led to the edge. Figure in right shows the feedback solution on this simple graph by red arrows that maps each node to the corresponding optimal local planner. As it is seen, the FIRM solution drives the system to more informative regions in the space, with less collision probabilities.

Related Publications

  • Ali-akbar Agha-mohammadi, Suman Chakravorty, Nancy M. Amato. "FIRM: Feedback Controller-Based Information-State Roadmap, A Framework for Motion Planning Under Uncertainty," In Proc. IEEE Int. Conf. Intel. Rob. Syst. (IROS11), San Francisco, CA, Sep 2011.
  • Ali-akbar Agha-mohammadi, Suman Chakravorty, Nancy M. Amato. "On the Probabilistic Completeness of the Sampling-based Feedback Motion Planners in Belief Space," In Proc. IEEE Int. Conf. Robot. Autom. (ICRA12), Saint Paul, Minnesota, May 2012.
  • Ali-akbar Agha-mohammadi, Suman Chakravorty, Nancy M. Amato. "Sampling-based Feedback Motion Planning Under Motion Uncertainty and Imperfect Measurements," Technical Report, TR11-007, Department of Computer Science and Engineering, Texas A&M University, Dec 2011
  • Ali-akbar Agha-mohammadi, Suman Chakravorty, Nancy M. Amato. "Feedback-based Information Roadmap for Nonholonomic Motion Planning via Dynamic Feedback Linearization," IEEE Int. Conf. Intel. Rob. Syst. (IROS12), Submitted.
  • Ali-akbar Agha-mohammadi, Suman Chakravorty, Nancy Amato. "Nonholonomic Motion Planning in Belief Space using Sampling-based Periodic Feedback Controllers", Workshop on the Algorithmic Foundations of Robotics (WAFR’12), Cambridge, Massachusetts, 2012, Submitted.

Links