Trustworthy Target Tracking with Collaborative Deep Reinforcement Learning in EdgeAI-Aided IoT

Abstract : In this paper, we propose DRL Track, a framework for target tracking with a collaborative deep reinforcement learning (C-DRL) in Edge-IoT with the aim to obtain high quality of tracking (QoT) and resource-efficient performance. In DRLTrack, a huge number of IoT devices are employed to collect data about a mobile target. One or two edge devices coordinate with the IoT devices and collaboratively detect the target by using the C-DRL approach and form an area around the target by a group of IoT devices. To maintain such an area during the tracking time, we employ a deep Q-network (DQN) to track the target from one group to another. An EdgeAI sitting on the top of the edge devices has the control of the C-DRL approach during tracking and can identify a sequence of tracks. We validate the performance of DRLTrack through simulations and it shows superior performance compared with existing work.

 EXISTING SYSTEM :

 ? Mobile edge computing can then complement the goals of the access network to solve existing challenges including quality of service/experience, security, and power consumption as part of the necessary network transformation. ? The coexistence of eMBB and URLLC with different service requirements is also a challenge. ? It is quite a challenge for 5G networks to meet URLLC specifications and this will entail major changes to the system architecture of the existing telecom infrastructure. ? Both architectures are designed to perform real-time analytics for traffic engineering and monitoring alongside existing protocols, such as BGP.

 DISADVANTAGE :

 ? One critical problem in MEC is to decide which edge servers should their computing tasks be offloaded to. ? The tradeoff between the communication overhead and the computing capability increases the complexity of the server assignment problem. ? To achieve this objective, we firstly formulate a task partition and scheduling optimization problem, which allows all received tasks in the network to be executed with minimized latency given the offloading strategy. ? A heuristic task partition and scheduling approach is developed to obtain a near-optimal solution of the non-convex integer problem.

 PROPOSED SYSTEM :

 • The mobile edge computing supports different network infrastructures including those proposed by 3GPP for 5G in particular. • ETSI proposes a platform that creates a multi-access edge system, which uses several heterogeneous access technologies, such as those proposed by 3GPP, and local or external networks, among others. • The mobile edge computing and NFV architectures proposed by ETSI are complementary. • In, the authors propose an effective approach for collecting globally available resource information through a mobile network architecture based on the SDN.

 ADVANTAGE :

 ? We investigate the task partition and scheduling under the collaborative computing framework, in which the adjustment on workload allocation for a task can affect the performance of other tasks, which makes the problem more complex. ? We evaluate the performance of the proposed AI-based collaborative computing approach in a vehicular network simulated by VISSIM, where TPSA is applied to schedule computing tasks according to the policy given by the DDPG algorithm. ? The proposed TPSA algorithm can achieve a performance very close to the brute-force scheme. ? The proposed TPSA algorithm can achieve a near-optimal performance for task partition and scheduling with low computation complexity.