PRK-Based Scheduling for Predictable Link Reliability in Wireless Networked Sensing and Control Hongwei Zhang, Xiaohui Liu , Chuan Li , Yu Chen , Xin Che , Feng Lin*, Le Yi Wang*, George Yin Department of Computer Science, Wayne State University, Detroit, Michigan, {hongwei,xiaohui,chuan,yu_chen}@wayne.edu *Department of Electrical and Computer Engineering, Wayne State University, Detroit, Michigan, {flin,lywang}@wayne.edu Department of Mathematics, Wayne State University, Detroit, Michigan, gyin@wayne.edu From Open-loop Sensing to Closed-loop Sensing and Control Smart grid: From centralized generation to distributed generation Grand societal challenges Power grid With ~2,459 million metric tons of CO2 emission per year, electricity generation accounts for ~41% of USA’s total CO2 emission Over 60% of today’s energy is wasted during distribution Transportation Car accidents cause over 1.4 million fatalities and 50 million injuries per year across the world Motor vehicles account for >20% of the world’s energy use and >60% of the world’s ozone pollution Connected Vehicles From passive to active safety: lane departure warning, collision avoidance From single-vehicle control to platoon control & integrated infrastructurevehicle control: networked fuel economy and emission control From wired intra-vehicle networks to wireless intra-vehicle networks Multiple controller-area-networks (CANs) inside vehicles • • • 50+ kg of wires increased, reduced fuel efficiency Lack of scalability: hundreds of sensors, controllers, and actuators Wiring unreliability: warranty cost, reduced safety Control-Oriented Wireless Networking: Physical-Ratio-K (PRK) Model Ratio-K-based scheduling is highly sensitive to the choice of K 200 150 Median PDR gain Median throughput gain Physical-Ratio-K (PRK) Interference Model Key idea: use link reliability requirement as the basis of instantiating the ratio-K model Model: given a transmission from node S to node R, a concurrent transmitter C does not interfere with the reception at R iff. P( S , R) P(C , R) K S , R ,TS ,R Optimality of PRK-Based Scheduling 100 25 0 -50 -100 -5 P( S , R) K S , R ,TS ,R S R C Challenges of PRK-Based Scheduling On-the-fly instantiation of the PRK model parameter K S ,R,T Dynamics and uncertainties in application requirements as well as network and environmental conditions Protocol signaling in the presence of large interference range as well as anisotropic, asymmetric, and probabilistic wireless communication S ,R 50 0 k 5 Highest throughput is usually achieved at a K less than the minimum K for ensuring a certain min. link reliability, and this is especially the case when link reliability requirement is high (e.g., for mission-critical sensing and control) Throughput loss(%) Behavior of Ratio-K-Based Scheduling Possible performance gain (%) Co-channel interference as a major obstacle for predictable reliability, real-time, and throughput in wireless networking Reliability as low as ~30% in current wireless scheduling/MAC protocols, thus not suitable for real-time, safety-critical networked control Despite decades of research and practice, high-fidelity interference models that are suitable for distributed, field-deployable protocol design are still missing Ratio-K model (i.e., protocol model) is local but not of high-fidelity SINR model (i.e., physical model) is of high-fidelity but non-local 20 15 10 5 0 10 20 30 40 50 60 70 80 PDR requirement(%) 90 95 99 Throughput loss is small, and it tends to decrease as the PDR requirement increases Distributed PRK-Based Scheduling for Predictable Link Reliability Minimum-variance regulation controller Protocol signaling via local signal maps PRK model instantiation: The control input that minimizes var[y(t 1)] As minimum-variance regulation control Local signal map: maintains wireless while ensuring E[ y (t 1)] TS ,R is signal power attenuation between nodes Basic problem formulation close-by cy (t ) (1 c)YS , R (t ) TS , R Reference input: desired link reliability TS , R I R (t ) U (t ) (1 c)a (t ) Simple approach to online estimation of Control output: actual link reliability YS , R wireless signal power attenuation var [y(t 1)] and the minimum value of is Control input: PRK model parameter K S ,R,T PC , R Ptotal PI y2,min (t 1) (1 c)2 a(t )2 U2 (t ) power loss Ptx PC , R 2 Interference from outside exclusion where U (t ) and U (t ) are the mean and variance of the region treated as disturbance changes of interferen ce from outside the exclusion region. PRKS: architecture of PRK-based scheduling S ,R • Minimize variance of YS , R while ensuring its mean value of TS , R Predictable link reliability in PRKS Challenge: Difficult to identify closed-form relation between control input and control output Refined control problem formulation Leverage communication theory result on the relation between YS , R and receiver-side SINR (i.e., PS , R I R ) YS , R f PS , R I R Convergence of distributed controllers From I R (t ) to K S , R,TS ,R (t 1) “Desired change in receiver-side interference I R ” as control input Comparison with existing protocols Linearization of the non-linear f(.) YS , R (t ) a(t )( PS , R (t ) I R (t )) b(t ) where a(t ) f ' ( PS , R (t ) I R (t )) b(t ) f ( PS , R (t ) I R (t )) ( PS , R (t ) I R (t )) f ' ( PS , R (t ) I R (t )) Larger networks