Cross-Layer Adaptive Feedback Scheduling of Wireless Control System Presented by Bin Tang
Wireless Control Systems (WCSs) Spatially distributed nodes of sensors, controllers, and/or actuators interconnected with wireless links Advantages Flexible installation and maintenance, mobile operation Hazardous and inaccessible environment Cheaper cost Disadvantages Path loss, multi-path fading, interference, etc Delay, packet losses, jitters get more pronounced
WCS System Model Independent control loop Smart sensor (S) Smart actuator (A) Controller (C) Physical process (P) IEEE801.11b Variable channel capacity Control performance is decided by Deadline Miss Ratio (DMR)
Deadline Miss Ratio (DMR) Deadline is equal to the sampling period Two cases for DMR: Sample data/control command is lost through wireless medium (bit error, interference, varying strength) Control command is received by actuator, later than deadline
Cross-layer adaptive feedback scheduling (CLAFS) scheme Information exchanging b/w application layer and physical layer Dynamically adjust sampling period w.r.t. transmission rate deadline miss ratio (DMR) Feedback scheduler uses PID algorithm with adaptive parameters Event-driven invocation algorithm
Sampling Period Sampling period of the control loop QoC and workload Delay and DMR Adjust sampling period based on transmission rate and DMR
Analysis of DMR over WLAN DMR at appropriate non-zero desired Smaller sample period, larger DMR Larger sample period, poorer Quality of Control
PID (Proportional-Integral- Derivative) control algorithm
feedback scheduling Sampling period at j th invocation instant of the feedback scheduler: h(j) = min{h(j-1)+Δh(j), h max } h max : maximum allowable sampling period Δh(j)=K*e(j) e(j)=ρ(j)-ρ r : actual and setpoint DMR K: proportional coefficient
Proportional Coefficient K K 0 : from simulation Δρ +, Δρ - : parameter
Optimal Quality of Control: (h r, ρ r )
Event-triggered invocation Time-triggered not suitable for wireless A Execution-request event iff |ρ(j)-ρ r |≥δ
Performance Evaluation Matlab with TrueTime toolbox Two scenarios: Scenario I: Controller and process close to each other; 11Mbps: no interfering Scenario I: Increased distance; 5.5 Mbps; interfering signal
Wireless Sensor/Actuator Network Design for Mobile Control Applications
Automated Architecture Semi-Automated Architecture
Experimental Analysis of Link Quality
Dealing with Packet Loss on Actuator Idea: when sensor data lost, actuator still produce control command Prediction from prev commands û(k): estimate of k-th control command
Inverted pendulum system: PID control algorithm r(k): desired system output y(k): measured system output Integral of absolute error (IAE)
Random Thoughts Control Side: How to find the optimal operating point (h r, ρ r )? So far reactive, not proactive Automated or semi-automated? Wireless Network Side: Can MAC layers and routing algorithms play a more active in WCS? Sensors/actuators/controllers coordination is not considered Data sensed by sensor is solely for the purpose of control Control application and other applications together