380 likes | 535 Views
Integrated Play-Back, Sensing, and Networked Control. Vincenzo Liberatore Division of Computer Science. Research supported in part by NSF CCR-0329910, Department of Commerce TOP 39-60-04003, NASA NNC04AA12A, and an OhioICE training grant. Networked Control. Computing in the physical world
E N D
Integrated Play-Back, Sensing, andNetworked Control Vincenzo Liberatore Division of Computer Science Research supported in part by NSF CCR-0329910, Department of Commerce TOP 39-60-04003, NASA NNC04AA12A, and an OhioICE training grant.
Networked Control • Computing in the physical world • Components • Sensors, actuators • Controllers • Networks Control Playback
Networked Control • Enables • Industrial automation [BL04] • Distributed instrumentation [ACRKNL03] • Unmanned vehicles [LNB03] • Home robotics [NNL02] • Distributed virtual environments [LCCK05] • Power distribution [P05] • Building structure control [SLT05] • Merge cyber- and physical- worlds • Networked control and tele-epistemology [G01] • Sensor networks • Not necessarily wireless or energy constrained • One component of sense-actuator networks Control Playback
Flow Sensor data Remote controller Control packets Timely delivery Stability Safety Performance Information Flow Control Playback
S&R Tele-operation Autonomy Autonomy • S&R and real-time • Autonomy • Hide networked RT • Hard to build a fully reliable system • Tele-operation • Network non-determinism is serious problem • S&R • Reduce time constants • Especially important for unexpected occurrences [NLN02] Control Playback
Networked Evaluation [EESR 2005] Control Playback
Stability (and safety) Objective Remote controller makes unstable system stable Extensive research [Z01] and references therein Problem Errors, network partitions, failures make stability impossible Tracking Objective The S&R system should do what it is supposed to In spite of network non-determinism (failures, security, etc.) Problem Benchmarks (NIST?) Disturbance cancellation Objective The S&R system should do what it is supposed to do In spite of network non-determinism and uncertainty in the environment Way out Use simple tasks Scalability [L04] Number of nodes Space networks? “Geographic” Administrative Functional Conclusion RT S&R benchmarks needed! Metrics Control Playback
Methodology (I): Co-Simulation Control Playback [BLP03, HLB05]
A Modest Proposal • Application benchmark • National Lambda Rail • “NLR is planned to be capable of supporting both production and experimental networks. • Not a single network or a single test bed but facilities to build multiple networks and multiple test beds at all of layers 1-3 including optical, switched, and routed. • Goal is to have both persistent and flexible infrastructure(s) • Foster network research” • Support QoS • Real-Time Overlay • Support end-to-end RT S&R Control Playback
Playback Buffers [Infocom 2006] Control Playback
Playback Buffers • Play-back buffers • Main objective • Smooths out network non-determinism • Multimedia buffers • Important source of inspiration • Physics versus multimedia quality • Playback delay computed in advance • Affects control signal computation • Round-Trip Times • TCP RTO • Another source of inspiration • Large time-out cost Control Playback
Algorithm Control Playback
Main Ideas • Predictable application time • If control applied early, plant is not in the state for which the control was meant • If control applied for too long, plant no longer in desired state • Keep plant simple • Low space requirements • Integrate Playback, Sampling, and Control Control Playback
Algorithm • Send regular control • Playback time • Late playback okay • Expiration • Piggyback contingency control Control Playback
X X Deadwood packets • Old • Received after the expiration time • Out-of-order • Later control more appropriate for current plant state • Would get us into a deadlock • New packet resets the playback timer • Keep resetting until no signal applied • “Quashed” packet • Discard! controller plant Playback delay Control Playback
Countermand control • Scenario • Packet i+1 overtakes packet I • ti+1 << ti • Likely caused by delay spike • New signal countermands previous one controller plant ti Playback delay ti+1 Control Playback
Playback delays • Modular component • Compute playback delay t and sampling period T • Use short term peak-hopper [EL04] • Original peak-hopper for TCP RTO • Too conservative for networked control • Aggressively attempt to decrease t • Aggressively attempt to decrease T • Add upper bound on playback delay t • Avoid dropping deadlock packets • Bound t ≤ T+RTT • Caps t and T • Must estimate lower-bound on RTT • Use symmetric of peak-hopper • Add negative variability estimate to compensate for short-term memory Control Playback
Playback Delays (I) Calculate current RTT variability Positive variability coefficient Negative variability coefficient if then Update min RTT estimate Age min RTT estimate Calculate Control Playback
Playback Delays (II) if then Attempt to avoid quashed packets else Increase sampling period Control Playback
Control Pipes • Bandwidth and delays • t is playback delay • T is sampling period • 1/T proportional to bandwidth • Control pipe • T«t • Multiple in-flight packets • Pipe depth • Bound by constraint t ≤ T+RTT • Keep pipe predictable Control Playback
Observer • Estimate future plant state • Plant sample current state, including local variables • Keep log of outstanding control packets • Assumption on packet delivery • Future packet delivery is uncertain • Purge from log • Old packets • Packet that should be overtaken by new control • Countermands signals generated when delay spike is transient • Out-of-order packets Control Playback
Evaluation Control Playback
Network Model • Simulated network • Losses: Gilbert model • Delays • Shifted Gamma distribution • Heavy tail • Low probability of out-of-order delivery • Correlate delays to introduce delay spikes • Wide-area implementation • Use RT scheduling whenever possible • Use otherwise unloaded machines • RT made little difference • Host worldwide, heterogeneous conditions Control Playback
Plant • Scalar linear plant • Plant state x(t) • Input u(t) (control) • Output y(t) • Disturbances v(t), w(t) • Akin to white noise • Deadbeat controller • Aggressive Control Playback
Metrics • Metrics • Root-mean square output • Output: 99-percentile • Comparison • Open-loop plant u(t)=0 • Proportional controller (no buffer) • Proportional controller with constant delays Control Playback
Plant output Open Loop Play-back Control Playback
Packet losses Figure 8 Control Playback
Sampling period Root-mean-square error Imperfection of the control pipe Control Playback
Agent-Oriented S&R Software [WORDS 2003] Control Playback
Agent-oriented S&R software • Progress • Agent-oriented platform • Compliant control • Future work • Application-oriented middleware • E.g., Scheduling of mobility • AI (knowledge, planning, learning) • Security Control Playback
Virtual Robots: The Core GUI, interface Thin-legacy layer On-board controllers Agent types Control Playback
Relationship: Virtual inclusion Hierarchical organization Chain of command Control Playback
Virtual Containment • Analogy • A robotic platoon “contains” individual robot • Not necessarily in terms of ontology • Application • Task-oriented teams • Layering Control Playback
Parameter (range) Robot id Go!!! Methods Task-space: Fluid dynamics Experience Control Playback
Acknowledgments • Students • Ahmad al-Hammouri • David Rosas • Zakaria Al-Qudah • Huthaifa Al-Omari • Nathan Wedge • Qingbo Cai • Prayas Arora • Colleagues • Michael S. Branicky Control Playback
Conclusions (I) • Sense-and-Respond • Merge cyber-world and physical world • Critically depends on physical time • Playback buffers integrated with • Sampling (adaptive T) • Control (expiration times, performance metrics) • Packet losses • Reverts to open loop plant (contingency control) Control Playback
Conclusions (II) • Playback delay t • Adapts to network conditions • Sampling period T • Avoids imperfection of control pipe • Simulations and emulations • Low variability around set point • Robust Control Playback
Conclusions (III) • Remote supervision of robotic manipulation • Compliant control • Local encapsulation • Gentle, compliant, tolerant to network vagaries • Agent-based software • Hierarchical • Demonstration • Future work: middleware, AI, security http://home.case.edu/~vxl11/NetBots/ Control Playback