In this paper, new cooperative fault detection algorithms are developed to detect the inter-vehicle communication failures in a network of multiple vehicle systems. The proposed detection algorithms work based on the notion that communication failure can lead to breaks or delays in the exchanged messages. Depending on the communication topology and the communication devices employed, different algorithms are required to diagnose failure and identify the faulty vehicles in the group. Algorithms

more »

... re developed for the cases with communication that is bidirectional, unidirectional, and employed with separate transmitter and receiver units. For each algorithm, the necessary conditions on the communication graph topology, under which failure is detectable, are derived. Using probability analysis the reliability of the proposed algorithms is also investigated. 2 detection framework is developed to detect the joint position and velocity sensors faults. The position and velocity sensor faults are detected by analyzing the position and velocity constraints in connected joints of manipulators. If the difference between the measurement of one sensor and other sensors are larger than some threshold a failure is concluded. Recently, the practical implementation issues such as computation time, communication requirements, and model uncertainties in the field of cooperative vehicle system control have been considered widely. 11-17 Towards considering practical implementation issues, in this paper the communication fault detection over a network of cooperative vehicles operating in a decentralized fashion is considered. The communication failure can potentially lead to break/delay in the communicated messages. Some examples of the communication failures leading to large communication delays for the team of cooperative vehicles can be found in Refs. 11, [18] [19] , the wireless communication packet loss/delay is considered; also, in Ref. 11, the communication failure in the formation flight of multiple UAVs leads to break in the communicated messages that enforce the fleet to redefine the communication topology. In this paper, it is assumed that each vehicle is equipped with a high performance communication device and the cooperative vehicles share a communication channel 22 to communicate to each other. It is also assumed that in the normal conditions the high performance communication device enables vehicles to communicate with neighboring vehicles with a very small delay, typically smaller than the sampling time. Then, the faulty condition is defined as follows: Faulty condition: The high performance communication device of one vehicle in the team fails. It is desired to provide some cooperation among the neighboring vehicles to diagnose the defined communication failure. A few research works have addressed the communication failure detection for multiple vehicles. A very closely related work is presented in Ref. 11 where it is desired to manage the communication failures in formation flight of multiple UAVs; it is assumed that the communication failure leads to complete blockage of information flow to/from faulty vehicle. Then, in Ref. 11 to keep all aircrafts informed about all operational members in the group, it is suggested to use an extra broadcasting communication channel. If after some specific time one aircraft has not sent its "alive" signal through the backup communication channel, that aircraft is considered lost. In another related work, 18 two faults for formation flight of UAVs are considered: 1) GPS sensor failure and 2) wireless communication packet losses. To detect the GPS sensor failure a state/output observer is used which monitors the behavior of a UAV. The output of the observer is compared with the GPS data, and if the difference is larger than some threshold then a GPS fault is identified. Also, in Ref. 18 to detect the communication packet loss/delay, the faults are identified by numbering the packets sequentially and the number of the packet is also transmitted; a mismatch between the expected packet number and the received packet number implies the occurrence of a fault (packet loss). The communication failure detection of cooperative vehicles becomes more challenging when a decentralized structure is used, as every vehicle should rely on the local information. Also, depending on the choice of communication topology (unidirectional or bidirectional) or whether the failure happens to transmitter or receiver devices, different scenarios can happen. In this paper three main cases are considered: 1) unidirectional communication (directed topology) 2) bidirectional communication (undirected communication topology) 3) separate transmitter and/or receiver failures (unidirectional or bidirectional). The latter considers the cases where the failure does not necessarily apply to both receiver and transmitter devices. The failure diagnosis algorithm for each vehicle includes: 1) Monitoring and detecting the faulty situation, 2) Identifying the faulty vehicle in the team; i.e. each vehicle must determine if itself is faulty or its neighbors by cooperating with neighbors, it must also determine which neighbor is faulty.