Abstract
Smart and agile drones are fast becoming ubiquitous at the edge of the cloud. The usage of these drones is constrained by their limited power and compute capability. In this paper, we present a Transfer Learning (TL) based approach to reduce on-board computation required to train a deep neural network for autonomous navigation via value-based Deep Reinforcement Learning for a target algorithmic performance. A library of 3D realistic meta-environments is manually designed using Unreal Gaming Engine and the network is trained end-to-end. These trained meta-weights are then used as initializers to the network in a test environment and fine-tuned for the last few fully connected layers. Variation in drone dynamics and environmental characteristics is carried out to show robustness of the approach. Using NVIDIA GPU profiler, it was shown that the energy consumption and training latency is reduced by 3.7 $\times$ and 1.8 $\times$ respectively without significant degradation in the performance in terms of average distance traveled before crash i.e. Mean Safe Flight (MSF). The approach is also tested on a real environment using DJI Tello drone and similar results were reported. The code for the approach can be found on GitHub: https://github.com/aqeelanwar/DRLwithTL .
Highlights
Over the past decade, Unmanned Aerial Vehicles (UAVs) are emerging as a new form of IoT devices being used in varied applications such as reconnaissance, surveying, rescuing and mapping
Cost and power is taken into account, such systems are heavy, expensive and power hungry, making them almost impossible to be used in low cost Micro Aerial Vehicles (MAVs)
We propose a two-phase approach to the problems related to Deep Reinforcement learning (DRL) which combines offline and online learning using Transfer Learning and fine-tuning
Summary
Over the past decade, Unmanned Aerial Vehicles (UAVs) are emerging as a new form of IoT devices being used in varied applications such as reconnaissance, surveying, rescuing and mapping. The problem is, for RL related applications how can we implement a neural network training on resource-constrained edge nodes with reduced power and latency and without losing on performance. Reference [7] uses an approach where the network is trained on simulated environments posing RL as supervised learning problem and deployed on new unknown environments This transfer of knowledge without further fine-tuning doesn’t always work well and is tightly tied to the co-relation or similarity between the train and test environments. In this paper we show we can use Transfer learning, to segment a deep network into trainable and non-trainable part reducing the training computations, for underlying task without compromising too much on its performance. This reduction is computation directly translates to reduced training energy leading to an energy efficient system
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
