The timing of sensor-based control systems is crucial. Critical servo-level periodic tasks that fail to meet their deadlines result in losing data or missing control cycles. This can lead to a loss in performance in the best case, and can cause serious damage to equipment or human injury in the worst case. It is therefore critical that the timing of these systems is predictable and controllable. A dynamically reconfigurable system can change in time without the need to halt the system. Such systems may have many sensors or actuators, only a subset of which are used at any time. Alternately the same hardware is used in a different configuration. In this paper we propose the maximum-urgency-first algorithm, which can be used to predictably schedule dynamically changing systems. We show that it is a significant improvement over the rate monotonic algorithm, which can only be used to schedule static systems. The maximum-urgency-first scheduler has been implemented as the default scheduler of CHIMERA II, a real-time operating system being used to control sensor-based control systems both at Carnegie Mellon University and elsewhere.

Abstract

This paper addresses the problem of planning a task for a robotic system comprised of a manipulator mounted on a mobile base. The task planning problem is formulated as a nonlinear optimization problem. The cost of point-to-point motion in three-dimensional Cartesian space is decomposed into two components representing the qualitative difference between motion due to the mobile base and motion due to the manipulator system. Task specifications at each end of the motion impose constraints on the endpoint configurations. The resulting regions of feasible positions and configurations are unconnected and nonconvex. Thus, standard algorithms for nonlinear optimization lead to nonextremal solutions. We present a heuristic method for searching a tree of starting points for a standard numerical algorithm to find a global minimum for the cost function. The problem formulation is illustrated for a 3 degrees-of-freedom (DOF) manipulator on a simple 2 DOF mobile base and trade-offs between base motion and manipulator motion are evaluated with respect to cost function weighting coefficients.

Computing the exact maximum likelihood or maximum a posteriori estimate of the environment is computationally expensive in many practical distributed sensing settings. We argue that this computational difficulty can be overcome by increasing the number of sensor measurements. Based on our work on the connection between error correcting codes and sensor networks, we propose a new algorithm which extends the idea of sequential decoding used to decode convolutional codes to estimation in a sensor network. In a simulated distributed sensing application, this algorithm provides accurate estimates at a modest computational cost given a sufficient number of sensor measurements. Above a certain number of sensor measurements this algorithm exhibits a sharp transition in the number of steps it requires in order to converge, leading to the potentially counter-intuitive observation that the computational burden of estimation can be reduced by taking additional sensor measurements.

No abstract