As a new computing paradigm, mobile edge computing (MEC) enables users to execute computation-intensive tasks at the network edge nodes (ENs) through computation offloading. Energy consumption of computation offloading is envisioned as a significant metric to satisfy the high quality of experience (QoE). In multi-ENs MEC networks, computation scheduling and power control of each user is tightly coupled with task offloading. Moreover, due to the stochastic task arrivals and unstable wireless channels, it is challenge to allocate resource for efficient offloading without prior information of tasks and channels. In this paper, we propose an individualized utility metric of each user. We formulate the problem of computation scheduling and power control of each user as a stochastic optimization problem. We aim to maximize the long-term averaged utility quality of all users by jointly optimizing the computation scheduling, taskpartition factor and power control. We use Lyapunov optimization technique to convert the long-term stochastic problem into a series of deterministic sub-problems in each time slot. We propose an online algorithm for utility quality maximization (OAUQM). The asymptotic optimality and queue stability of our algorithm are analyzed. Experimental simulations are conducted to evaluate the performance of the proposed algorithm against the benchmark offloading algorithms in terms of utility quality and energy consumption.
Generative adversarial imitation learning (GAIL) has shown good results in several research areas by taking advantage of generative adversarial networks. However, GAIL lacks a reward mechanism and usually adopts a model-free approach based on stochastic policies, which is not ideal for solving complex, dynamically uncertain population intelligence problems, especially in the face of autonomous driving environments. In this paper, a policy framework is shaped by combining the human knowledge with GAIL (HKGAIL). HKGAIL embeds human decision models into the learning process to infer the underlying structure of expert demonstrations. The skills learned from expert demonstrations can directly guide the actions (policies) of the learning process of the agents, and the policies can be optimized through the feedback function of the discriminator. Experiments on both driving and landing tasks show that HKGAIL was able to better fit the policy close to the expert, and was 16.2% safer than GAIL for the driving task.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.