Vision-based UAV Guidance for Autonomous Landing with Deep Neural Networks

Bicer, Yunus; Moghadam, Majid; Sahin, Cagatay; Eroglu, Batuhan; Üre, Nazım Kemal

doi:10.2514/6.2019-0140

Cited by 20 publications

(12 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Early experiments showed a vertical offset between the simulated and real trajectories, caused by a slight inaccuracy of our motor thrust equation (6). We were able to compensate for this by increasing the hover PWM to 48000 during real testing.…”

Section: B Experiments Resultsmentioning

confidence: 90%

“…Furthermore, the same policy can be initiated from a wide variety of starting states. Results could further improve by additional system identification of the total motor thrust, where a slight mismatch might have been caused by making too strong assumptions about the scaling of the single motor model in (5) to the full motor model in (6). A perfect thrust model will however not exist, due to motor wear and tear and the strong relation of the drone's battery level to its thrust output, which is also reported in [1].…”

Section: B Experiments Resultsmentioning

confidence: 98%

“…Least-squares policy iteration (LSPI) was employed to autonomously land on a marker [10] and the deep deterministic policy gradient (DDPG) algorithm [16] was used to navigate a descending quadrotor to land on a moving platform [9]. Finally, the work in [6] involved a convolutional neural network to estimate the heading angle to aid UAV landing in the case of a sensor failure. However, none of the approaches considered inclined landing and none of the methods developed can be directly applied to this problem.…”

Section: Related Workmentioning

confidence: 99%

“…Recent advances in deep reinforcement learning (DRL) with continuous action spaces have made this approach suitable also for quadrotor control [1], [2], [3], [4], including landing controllers [5], [6], [7], [8], [9], [10]. However, no results have yet been reported for inclined landing.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Inclined Quadrotor Landing using Deep Reinforcement Learning

Kooi,

Babuška

2021

Preprint

View full text Add to dashboard Cite

Landing a quadrotor on an inclined surface is a challenging manoeuvre. The final state of any inclined landing trajectory is not an equilibrium, which precludes the use of most conventional control methods. We propose a deep reinforcement learning approach to design an autonomous landing controller for inclined surfaces. Using the proximal policy optimization (PPO) algorithm with sparse rewards and a tailored curriculum learning approach, a robust policy can be trained in simulation in less than 90 minutes on a standard laptop. The policy then directly runs on a real Crazyflie 2.1 quadrotor and successfully performs real inclined landings in a flying arena. A single policy evaluation takes approximately 2.5 ms, which makes it suitable for a future embedded implementation on the quadrotor.

show abstract

Section: B Experiments Resultsmentioning

confidence: 90%

Section: B Experiments Resultsmentioning

confidence: 98%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Inclined Quadrotor Landing using Deep Reinforcement Learning

Kooi,

Babuška

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Many important tasks of autonomous systems depend on powerful machinelearning based components like DNNs. Typical examples might include autonomous vision-based takeoff and landing [2], speech-based communication with Air Traffic Control [17], or a compact representation of the complex dynamic aircraft model.…”

Section: Introductionmentioning

confidence: 99%

A Framework for the Analysis of Deep Neural Networks in Autonomous Aerospace Applications using Bayesian Statistics

Schumann

2020

2020 IEEE Security and Privacy Workshops (SPW)

View full text Add to dashboard Cite

Deep Neural Networks (DNNs) are considered to be key components in many autonomous systems. Applications range from vision-based obstacle avoidance to intelligent/learning control and planning. Safety-critical applications as found in the aerospace domain require that the behavior of the DNN is validated and tested rigorously for safety of the autonomous system (AUS).In this paper, we present a framework to support testing of DNNs and the analysis of the network structure. Our framework employs techniques from statistical modeling and active learning to effectively generate test cases for DNN safety testing and performance analysis.We will present results of a case study on a physics-based Deep recurrent residual neural network (DR-RNN), which has been trained to emulate the aerodynamics behavior of a fixedwing aircraft.

show abstract

Onboard CNN-Based Processing for Target Detection and Autonomous Landing for MAVs

Cabrera-Ponce

Martínez-Carranza

2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Vision-based UAV Guidance for Autonomous Landing with Deep Neural Networks

Cited by 20 publications

References 27 publications

Inclined Quadrotor Landing using Deep Reinforcement Learning

Inclined Quadrotor Landing using Deep Reinforcement Learning

A Framework for the Analysis of Deep Neural Networks in Autonomous Aerospace Applications using Bayesian Statistics

Onboard CNN-Based Processing for Target Detection and Autonomous Landing for MAVs

Contact Info

Product

Resources

About