Navigation Among Movable Obstacles via Multi-Object Pushing Into Storage Zones

Ellis, Kirsty; Hadjivelichkov, Denis; Modugno, Valerio; Stoyanov, Danail; Kanoulas, Dimitrios

doi:10.1109/access.2022.3233765

Cited by 8 publications

(4 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, due to the sparsity of the demonstration value parameters, they cannot provide comprehensive guidance for reinforcement learning. In Bendikas et al (2023), tasks are recursively decomposed into a series of subtasks, and then the agent is initialized with the existing critic network parameters to guide the current actor, thus achieving the guiding effect of the network Q function. In Wang et al (2021b) and Wang et al (2022a), for the control of complex assembly tasks, imitation learning is used to initially learn the outline of the trajectory, and then its parameters are used for subsequent force control learning, resulting in effective assembly force control strategies.…”

Section: Related Work Trajectory Learning Methods Based On Imitation...mentioning

confidence: 99%

Human skill knowledge guided global trajectory policy reinforcement learning method

Zang,

Wang,

Zha

et al. 2024

Front. Neurorobot.

View full text Add to dashboard Cite

Traditional trajectory learning methods based on Imitation Learning (IL) only learn the existing trajectory knowledge from human demonstration. In this way, it can not adapt the trajectory knowledge to the task environment by interacting with the environment and fine-tuning the policy. To address this problem, a global trajectory learning method which combinines IL with Reinforcement Learning (RL) to adapt the knowledge policy to the environment is proposed. In this paper, IL is proposed to acquire basic trajectory skills, and then learns the agent will explore and exploit more policy which is applicable to the current environment by RL. The basic trajectory skills include the knowledge policy and the time stage information in the whole task space to help learn the time series of the trajectory, and are used to guide the subsequent RL process. Notably, neural networks are not used to model the action policy and the Q value of RL during the RL process. Instead, they are sampled and updated in the whole task space and then transferred to the networks after the RL process through Behavior Cloning (BC) to get continuous and smooth global trajectory policy. The feasibility and the effectiveness of the method was validated in a custom Gym environment of a flower drawing task. And then, we executed the learned policy in the real-world robot drawing experiment.

show abstract

Section: Related Work Trajectory Learning Methods Based On Imitation...mentioning

confidence: 99%

Human skill knowledge guided global trajectory policy reinforcement learning method

Zang,

Wang,

Zha

et al. 2024

Front. Neurorobot.

View full text Add to dashboard Cite

show abstract

“…Compared with the literature [23], transforming the problem of moving obstacles into a simple problem of stationary obstacles solves the problem of considering the complexity of moving obstacles models in the Gaussian BLF. Compared with the literature [24], we simplified the calculation of moving obstacles in this method. The problem of moving obstacles can be integrated into complex obstacle avoidance strategies.…”

Section: ∆𝑇 = 𝐾 𝑇 (𝑅 𝑜 + 𝐿)/𝑢 𝑚𝑎𝑥mentioning

confidence: 99%

“…Cong [23] presented a method for obstacle avoidance utilizing reinforcement learning, implementing Qtable values in real-world mobile robots to perform tasks within actual scenarios. Ellis [24] proposed a technique for pushing multiple objects into designated storage areas, achieving obstacle avoidance for mobile robots in environments with moving obstacles by relocating them to predefined locations. However, these existing methods can be cumbersome and challenging to integrate with the Barrier Lyapunov Function (BLF) obstacle avoidance control strategy employed in our research.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Obstacle Avoidance Control of the AGV With Actuator Dead Zones Based on Fixed-Time Disturbance Observers and Command Filtering

Zhang,

Ding,

Chen

2024

IEEE Access

View full text Add to dashboard Cite

This study explores an obstacle avoidance control approach that takes into account both actuator dead zones and unpredictable disturbances in environments with moving obstacles. Initially, we enhance a Barrier Lyapunov Function (BLF) by introducing the concept of an equivalent safe distance. This innovation broadens the safety perimeter around obstacles by considering their velocity and movement duration, thereby simplifying the intricacies associated with moving obstacle models in avoidance strategies. Furthermore, we refine the auxiliary function within the obstacle avoidance control laws by incorporating exponential and saturation functions. This refinement ensures that as the directional error increases, the velocity of the Automated Guided Vehicle (AGV) decreases. Similarly, it guarantees a reduction in speed as the AGV nears obstacles, resulting in a smoother auxiliary function that prevents excessive speeding when approaching obstacles. Subsequently, we propose an obstacle avoidance control method for AGVs with dead zones, leveraging the enhanced BLF alongside the introduction of a fixed-time disturbance observer. This integration facilitates the estimation of disturbances within a predetermined timeframe. Unlike existing fixedtime disturbance observers, our method incorporates command filtering to mitigate the risk of computational overload. Ultimately, the efficacy of our proposed approach is validated through rigorous experimental simulations conducted on the Matlab platform.

show abstract

ViT-A: Legged Robot Path Planning using Vision Transformer A

Liu,

Lyu,

Hadjivelichkov

et al. 2023

2023 IEEE-RAS 22nd International Conference on Humanoid Robots (Humanoids)

View full text Add to dashboard Cite

Navigation Among Movable Obstacles via Multi-Object Pushing Into Storage Zones

Cited by 8 publications

References 29 publications

Human skill knowledge guided global trajectory policy reinforcement learning method

Human skill knowledge guided global trajectory policy reinforcement learning method

Obstacle Avoidance Control of the AGV With Actuator Dead Zones Based on Fixed-Time Disturbance Observers and Command Filtering

ViT-A: Legged Robot Path Planning using Vision Transformer A

Contact Info

Product

Resources

About

Navigation Among Movable Obstacles via Multi-Object Pushing Into Storage Zones

Cited by 8 publications

References 29 publications

Human skill knowledge guided global trajectory policy reinforcement learning method

Human skill knowledge guided global trajectory policy reinforcement learning method

Obstacle Avoidance Control of the AGV With Actuator Dead Zones Based on Fixed-Time Disturbance Observers and Command Filtering

ViT-A*: Legged Robot Path Planning using Vision Transformer A*

Contact Info

Product

Resources

About

ViT-A: Legged Robot Path Planning using Vision Transformer A