Air Learning: a deep reinforcement learning gym for autonomous aerial robot visual navigation

Boroujerdian, Behzad; Fu, William; Faust, Aleksandra; Reddi, Vijay Janapa

doi:10.1007/s10994-021-06006-6

Cited by 27 publications

(18 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…AI plays a vital role in improving the performance of robots including UAV in many applications [18][19][20]. The authors of [21] introduce the air learning open-source simulator environment for UAVs using Deep Reinforcement Learning (DRL) techniques. Federated Learning (FL) was used for improving UAV swarm computing scheduling and power allocation without exchanging sensitive raw data [22].…”

Section: Related Workmentioning

confidence: 99%

UAV Computing-Assisted Search and Rescue Mission Framework for Disaster and Harsh Environment Mitigation

Alsamhi

Shvetsov

Kumar

et al. 2022

Drones

132

View full text Add to dashboard Cite

Disasters are crisis circumstances that put human life in jeopardy. During disasters, public communication infrastructure is particularly damaged, obstructing Search And Rescue (SAR) efforts, and it takes significant time and effort to re-establish functioning communication infrastructure. SAR is a critical component of mitigating human and environmental risks in disasters and harsh environments. As a result, there is an urgent need to construct communication networks swiftly to help SAR efforts exchange emergency data. UAV technology has the potential to provide key solutions to mitigate such disaster situations. UAVs can be used to provide an adaptable and reliable emergency communication backbone and to resolve major issues in disasters for SAR operations. In this paper, we evaluate the network performance of UAV-assisted intelligent edge computing to expedite SAR missions and functionality, as this technology can be deployed within a short time and can help to rescue most people during a disaster. We have considered network parameters such as delay, throughput, and traffic sent and received, as well as path loss for the proposed network. It is also demonstrated that with the proposed parameter optimization, network performance improves significantly, eventually leading to far more efficient SAR missions in disasters and harsh environments.

show abstract

Section: Related Workmentioning

confidence: 99%

UAV Computing-Assisted Search and Rescue Mission Framework for Disaster and Harsh Environment Mitigation

Alsamhi

Shvetsov

Kumar

et al. 2022

Drones

132

View full text Add to dashboard Cite

show abstract

“…Luo et al (2021) investigate policies for industrial assembly with RL and imitation learning, tackling the prohibitively large design space. Krishnan et al (2021) introduce a simulator for resource-constrained autonomous aerial robots. Wurman et al (2022) develop automobile racing agent in simulation, the PlayStation game Gran Turismo, to win the world's best e-sports drivers.…”

Section: Roboticsmentioning

confidence: 99%

Reinforcement Learning in Practice: Opportunities and Challenges

Li¹

2022

Preprint

View full text Add to dashboard Cite

This article is a gentle discussion about the field of reinforcement learning for real life, about opportunities and challenges, with perspectives and without technical details, touching a broad range of topics. The article is based on both historical and recent research papers, surveys, tutorials, talks, blogs, and books. Various groups of readers, like researchers, engineers, students, managers, investors, officers, and people wanting to know more about the field, may find the article interesting.In this article, we first give a brief introduction to reinforcement learning (RL), and its relationship with deep learning, machine learning and AI. Then we discuss opportunities of RL, in particular, applications in products and services, games, recommender systems, robotics, transportation, economics and finance, healthcare, education, combinatorial optimization, computer systems, and science and engineering. The we discuss challenges, in particular, 1) foundation, 2) representation, 3) reward, 4) model, simulation, planning, and benchmarks, 5) learning to learn a.k.a. meta-learning, 6) off-policy/offline learning, 7) software development and deployment, 8) business perspectives, and 9) more challenges. We conclude with a discussion, attempting to answer: "Why has RL not been widely adopted in practice yet?" and "When is RL helpful?".

show abstract

“…As these accelerators get deployed at all computing scales, Figure 3: Different components in AutoSoC. In the front-end, we have Air Learning (Krishnan et al, 2021), which is used to perform task-algorithm-policy exploration. Once a policy is determined, it will be fed to the FlexACL block, which will take the neural network policy and generate a synthesizable RTL accelerator template to meet the performance and power specifications.…”

Section: Reinforcement Learningmentioning

confidence: 99%

“…AutoSoC uses Air Learning (Krishnan et al, 2021) as the robot simulator to train E2E models for aerial robot navigation.…”

Section: Air Learningmentioning

confidence: 99%

AutoSoC: Automating Algorithm-SOC Co-design for Aerial Robots

Krishnan,

Tambe,

Wan

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Aerial autonomous machines (Drones) has a plethora of promising applications and use cases. While the popularity of these autonomous machines continues to grow, there are many challenges, such as endurance and agility, that could hinder the practical deployment of these machines. The closed-loop control frequency must be high to achieve high agility. However, given the resource-constrained nature of the aerial robot, achieving high control loop frequency is hugely challenging and requires careful co-design of algorithm and onboard computer. Such an effort requires infrastructures that bridge various domains, namely robotics, machine learning, and system architecture design. To that end, we present AutoSoC, a framework for co-designing algorithms as well as hardware accelerator systems for end-to-end learning-based aerial autonomous machines. We demonstrate the efficacy of the framework by training an obstacle avoidance algorithm for aerial robots to navigate in a densely cluttered environment. For the best performing algorithm, our framework generates various accelerator design candidates with varying performance, area, and power consumption. The framework also runs the ASIC flow of place and route and generates a layout of the floor-planed accelerator, which can be used to tape-out the final hardware chip.

show abstract

Air Learning: a deep reinforcement learning gym for autonomous aerial robot visual navigation

Cited by 27 publications

References 36 publications

UAV Computing-Assisted Search and Rescue Mission Framework for Disaster and Harsh Environment Mitigation

UAV Computing-Assisted Search and Rescue Mission Framework for Disaster and Harsh Environment Mitigation

Reinforcement Learning in Practice: Opportunities and Challenges

AutoSoC: Automating Algorithm-SOC Co-design for Aerial Robots

Contact Info

Product

Resources

About