Presentation on theme: "Design of a reliable communication system for grid-style traffic light networks Junghoon Lee Dept. of Computer science and statistics Jeju National University."— Presentation transcript:
Design of a reliable communication system for grid-style traffic light networks Junghoon Lee Dept. of Computer science and statistics Jeju National University Rep. of Korea Song Han, Aloysius K. Mok Dept. of Computer sciences University of Texas Austin, Texas, USA
Performance Evaluation Background and Related Work Conclusion Introduction Channel schedule and routing Contents
Feedback control loop consists of geographically distributed components. needs reliable and timely communication. can install wireless process control protocol. WirelessHART standard has Clear Channel Assessment and provides predictable network access. ControllerActuators Sensors Controlled process Current state Current state
Controlled process Jeju area is located in the southernmost part of Korea. It hosts many pilot projects for the purpose of testing a system before its deployment. Vehicular telematics system Smart grid system
Traffic light network The traffic light is secure place for wireless nodes. Traffic lights form a grid topology in urban area. Process control applications can run on this network. Global control system Level 1 network Level 2 network
Performance Evaluation Routing and Scheduling Scheme Conclusion Introduction Background and Related Work Contents
WirelessHART standard IEEE GHz radioband physical link supports mesh networking directly. 16 frequency channels with 5 MHz guard time Channel blacklisting and hopping The slot-based access scheme with time synchronization The size of a single time slot : 10 ms central network manager schedule provisioning and updating CCA (Clear Channel Assessment) to avoid interference
Slot organization Setting the channel each time slot begins. 10 ms slots have CCA (Channel Condition Assessment). CCA and channel tuning just take several bit time. additional CCA and channel tuning for a slot
Protocol stack and mesh From Jianping Song et al.s paper in RTAS 2008 
Performance Evaluation Background and Related Work Conclusion Introduction Routing and Scheduling Scheme Contents
Motivation The standard does not define what to do when the CCA result is not good. If the CCA result is not good, can it be possible to take another route? Primary and secondary receivers can solve this problem. Two receivers wait for the message and the sender selects the receiver (channel) according to CCA result. The transmission paths can be easily split and merged over the grid topology.
System Model Traffic Light Network The traffic lights, installed in the intersections, form a grid network in the Manhattan-style road network. Each node can exchange messages directly with its vertical and horizontal neighbors, but not its diagonal one, considering the directional antenna. Control Loop reading state variables from sensors, deciding the control action, and sending the value of control variables to the actuators, : the network and the communication schedules
System Model The controller node is located at the left-top. For N 0,0 -> N 1,1, H 0,1 ->V 1,1 and V 1,0, H 1,1 paths are available. N 0,0 performs split and N 1,1 runs merge op. over 2 slots. CCA result is always right. Controller node Controller node Four 4 4 grids
Sample slot allocation table Split op. Merge op. slot allocation for a 3 3 grid and downlink.
Route decision Each rectangle can be considered to be a virtual link. Run the shortest path algorithm. The virtual link can be mapped to 2 slot transmissions. [….S i, j,......] Primary [ …, H i,j+1,V i+1,j+ 1,…] Secondary [ …, V i+1,j,H i+1,j+ 1,…] F1 substitution [….S i, j,......] Primary [ …, V i+1,j,H i+1,j+ 1,…] Secondary [ …, H i,j+1,V i+1,j+ 1,…] F2 substitution
Virtual link model Success probabilities for the two paths in a rectangle, F 1 and F 2 Select the bigger of the two as the primary path Set the error rate of the virtual link (1 – F 1 ) or (1- F 2 ) Run the Dijkstras shortest path algorithm
Routing and Scheduling Scheme Background and Related Work Conclusion Introduction Performance Evaluation Contents
Performance Evaluation Simulation Environment using SMPL which provides discrete event scheduling For simplicity, only the downlink graph was considered. 500 sets of link error rates are generated. The success ratios are averaged. Main metrics success ratio according to slot error rate, hop count. success ratio for the different routing schemes, dimension additional receive ratio effect of node failures
Performance Evaluation Each link has the same error rate Guilbert-Elliot error model End-to-end messages to each node a round Overall success ratio a large performance gap on the high slot error rate 7.9 % Effect of slot error rate
Performance Evaluation 11.4 % 7.8 % 5.5 % Success ratio classified by hop counts hop by hop success ratio no improvement for the 1 hop nodes The 6 hop node has 3 split-merge operations. SM shows stable success ratio for the hop count change.
Performance Evaluation Improvement by the routing scheme 6.69 % Each link has its own error rate. Enhance performance by the routing scheme based on the virtual link model just 3.6 % loop length overhead
Performance Evaluation (4/4) Effect of grid dimension 9.0 % average slot error rate is set to 0.1. A larger grid has more rectangles. Average hop length increases. Performance gap increases for a larger grid.
Performance Evaluation (4/4) Overhead analysis Additional receive due to the split operation total slots needed in a control round All nodes in the same row and column have the same SM operations
Performance Evaluation (4/4) Gap for diagonal nodes and other nodes Compare with a link disjoint path Each link has the same error rate Difference in average success ratio is less than 6 % For the diagonal case, the difference reaches up to 35 %
Performance Evaluation Routing and Scheduling Scheme Background & Related Work Introduction Conclusion
Conclusions Summary Process control for grid topology traffic light network Attempts two paths according to CCA by split-merge SM operation can be modeled as a virtual link Shortest path-based routing and slot allocation Enhances delivery ratio by up to 7.9 % on average Future Work Apply the split-merge operation for data collection and multicasting
Performance Evaluation (4/4) Effect of node failures 1 node failure makes 4 links unconnected A non-fringe node failure makes it difficult to run the SM operation. outperforms grid for all node failure range. 4.7 %