| 1. | Sashank Tirumala; Sagar Gubbi; Kartik Paigwar; Aditya Sagi; Ashish Joglekar; Shalabh Bhatnagar; Ashitava Ghosal; Bharadwaj Amrutur; Shishir, Kolathaya Learning Stable Manoeuvres for Quadruped Robots from Expert Demonstrations Conference 29th IEEE International Conference on Robot and Human Interactive Communication (RO-MAN), 2020. Abstract | BibTeX | Links:  @conference{@Shashank,
title = {Learning Stable Manoeuvres for Quadruped Robots from Expert Demonstrations},
author = {Sashank Tirumala; Sagar Gubbi; Kartik Paigwar; Aditya Sagi; Ashish Joglekar; Shalabh Bhatnagar; Ashitava Ghosal; Bharadwaj Amrutur; Shishir, Kolathaya},
url = {https://cps.iisc.ac.in/wp-content/uploads/2020/10/Learning-Stable-Manoeuvres-in-Quadruped-Robots-from-Expert-Demonstrations.pdf},
year = {2020},
date = {2020-07-28},
booktitle = {29th IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
abstract = {With the research into development of quadruped robots picking up pace, learning based techniques are being explored for developing locomotion controllers for such robots. A key problem is to generate leg trajectories for continuously varying target linear and angular velocities, in a stable manner. In this paper, we propose a two pronged approach to address this problem. First, multiple simpler policies are trained to generate trajectories for a discrete set of target velocities and turning radius. These policies are then augmented using a higher level neural network for handling the transition between the learned trajectories. Specifically, we develop a neural network-based filter that takes in target velocity, radius and transforms them into new commands that enable smooth transitions to the new trajectory. This transformation is achieved by learning from expert demonstrations. An application of this is the transformation of a novice user's input into an expert user's input, thereby ensuring stable manoeuvres regardless of the user's experience. Training our proposed architecture requires much less expert demonstrations compared to standard neural network architectures. Finally, we demonstrate experimentally these results in the in-house quadruped Stoch 2. },
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
With the research into development of quadruped robots picking up pace, learning based techniques are being explored for developing locomotion controllers for such robots. A key problem is to generate leg trajectories for continuously varying target linear and angular velocities, in a stable manner. In this paper, we propose a two pronged approach to address this problem. First, multiple simpler policies are trained to generate trajectories for a discrete set of target velocities and turning radius. These policies are then augmented using a higher level neural network for handling the transition between the learned trajectories. Specifically, we develop a neural network-based filter that takes in target velocity, radius and transforms them into new commands that enable smooth transitions to the new trajectory. This transformation is achieved by learning from expert demonstrations. An application of this is the transformation of a novice user's input into an expert user's input, thereby ensuring stable manoeuvres regardless of the user's experience. Training our proposed architecture requires much less expert demonstrations compared to standard neural network architectures. Finally, we demonstrate experimentally these results in the in-house quadruped Stoch 2. |
| 2. | Dholakiya, Dhaivat; Bhattacharya, Shounak; Gunalan, Ajay; Singla, Abhik; Bhatnagar, Shalabh; Amrutur, Bharadwaj; Ghosal, Ashitava; Kolathaya, Shishir Design, development and experimental realization of a quadrupedal research platform: Stoch Journal Article arXiv: Computer Science, 2019. Abstract | BibTeX | Links: @article{Dholakiya2019,
title = {Design, development and experimental realization of a quadrupedal research platform: Stoch},
author = {Dhaivat Dholakiya and Shounak Bhattacharya and Ajay Gunalan and Abhik Singla and Shalabh Bhatnagar and Bharadwaj Amrutur and Ashitava Ghosal and Shishir Kolathaya},
url = {http://www.rbccps.org/wp-content/uploads/2019/05/1901.00697.pdf},
year = {2019},
date = {2019-01-03},
journal = {arXiv: Computer Science},
abstract = {In this paper, we present a complete description of the hardware design and control architecture of our custom built quadruped robot, called the `Stoch'. Our goal is to realize a robust, modular, and a reliable quadrupedal platform, using which various locomotion behaviors are explored. This platform enables us to explore different research problems in legged locomotion, which use both traditional and learning based techniques. We discuss the merits and limitations of the platform in terms of exploitation of available behaviours, fast rapid prototyping, reproduction and repair. Towards the end, we will demonstrate trotting, bounding behaviors, and preliminary results in turning. In addition, we will also show various gait transitions i.e., trot-to-turn and trot-to-bound behaviors. },
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we present a complete description of the hardware design and control architecture of our custom built quadruped robot, called the `Stoch'. Our goal is to realize a robust, modular, and a reliable quadrupedal platform, using which various locomotion behaviors are explored. This platform enables us to explore different research problems in legged locomotion, which use both traditional and learning based techniques. We discuss the merits and limitations of the platform in terms of exploitation of available behaviours, fast rapid prototyping, reproduction and repair. Towards the end, we will demonstrate trotting, bounding behaviors, and preliminary results in turning. In addition, we will also show various gait transitions i.e., trot-to-turn and trot-to-bound behaviors. |
| 3. | Singla, Abhik; Bhattacharya, Shounak; Dholakiya, Dhaivat; Bhatnagar, Shalabh; Ghosal, Ashitava; Amrutur, Bharadwaj; Kolathaya, Shishir Realizing learned quadruped locomotion behaviors through kinematic motion primitives Conference 2019. Abstract | BibTeX | Links: @conference{Singla2018,
title = {Realizing learned quadruped locomotion behaviors through kinematic motion primitives},
author = {Abhik Singla and Shounak Bhattacharya and Dhaivat Dholakiya and Shalabh Bhatnagar and Ashitava Ghosal and Bharadwaj Amrutur and Shishir Kolathaya},
url = {http://www.rbccps.org/wp-content/uploads/2018/12/1810.03842.pdf},
year = {2019},
date = {2019-10-10},
journal = {ICRA 2019, Montreal},
abstract = {Humans and animals are believed to use a very minimal set of trajectories to perform a wide variety of tasks including walking. Our main objective in this paper is two fold 1) Obtain an effective tool to realize these basic motion patterns for quadrupedal walking, called the kinematic motion primitives (kMPs), via trajectories learned from deep reinforcement learning (D-RL) and 2) Realize a set of behaviors, namely trot, walk, gallop and bound from these kinematic motion primitives in our custom four legged robot, called the 'Stoch'. D-RL is a data driven approach, which has been shown to be very effective for realizing all kinds of robust locomotion behaviors, both in simulation and in experiment. On the other hand, kMPs are known to capture the underlying structure of walking and yield a set of derived behaviors. We first generate walking gaits from D-RL, which uses policy gradient based approaches. We then analyze the resulting walking by using principal component analysis. We observe that the kMPs extracted from PCA followed a similar pattern irrespective of the type of gaits generated. Leveraging on this underlying structure, we then realize walking in Stoch by a straightforward reconstruction of joint trajectories from kMPs. This type of methodology improves the transferability of these gaits to real hardware, lowers the computational overhead on-board, and also avoids multiple training iterations by generating a set of derived behaviors from a single learned gait. },
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Humans and animals are believed to use a very minimal set of trajectories to perform a wide variety of tasks including walking. Our main objective in this paper is two fold 1) Obtain an effective tool to realize these basic motion patterns for quadrupedal walking, called the kinematic motion primitives (kMPs), via trajectories learned from deep reinforcement learning (D-RL) and 2) Realize a set of behaviors, namely trot, walk, gallop and bound from these kinematic motion primitives in our custom four legged robot, called the 'Stoch'. D-RL is a data driven approach, which has been shown to be very effective for realizing all kinds of robust locomotion behaviors, both in simulation and in experiment. On the other hand, kMPs are known to capture the underlying structure of walking and yield a set of derived behaviors. We first generate walking gaits from D-RL, which uses policy gradient based approaches. We then analyze the resulting walking by using principal component analysis. We observe that the kMPs extracted from PCA followed a similar pattern irrespective of the type of gaits generated. Leveraging on this underlying structure, we then realize walking in Stoch by a straightforward reconstruction of joint trajectories from kMPs. This type of methodology improves the transferability of these gaits to real hardware, lowers the computational overhead on-board, and also avoids multiple training iterations by generating a set of derived behaviors from a single learned gait. |
| 4. | Ashwin, K P; Ghosal, Ashitava Mathematical model for pressure–deformation relationship of miniaturized McKibben actuators Conference Proceedings of the 3rd International and 18th National Conference on Machines and Mechanisms (iNaCoMM), 13.-15.12.17, Mumbai, 2018. Abstract | BibTeX | Links:  @conference{Ashwin2018,
title = {Mathematical model for pressure–deformation relationship of miniaturized McKibben actuators},
author = {K. P. Ashwin and Ashitava Ghosal},
doi = {10.1007/978-981-10-8597-0_23},
year = {2018},
date = {2018-08-29},
booktitle = {Proceedings of the 3rd International and 18th National Conference on Machines and Mechanisms (iNaCoMM), 13.-15.12.17, Mumbai},
journal = {Proceedings of the 3rd International and 18th National Conference on Machines and Mechanisms (iNaCoMM), 13.-15.12.17, Mumbai},
abstract = {A McKibben actuator/Pneumatic Artificial Muscle (PAM) is a soft actuator which has great potential in the field of bioinspired robotics. Miniaturized versions of PAMs or MPAMs of less than 1.5 mm diameter are ideal actuators for developing surgical devices due to their compliance and high power-to-weight ratio. Accurate mathematical model to represent the mechanics of PAM is an ongoing research. This paper develops a mathematical model which relates the input pressure to end-point deformation of a fabricated MPAM without external loading. The developed theoretical model is validated against experimental data for MPAM of lengths 60 and 70 mm. The model predicts the deformation of MPAM with standard error of less than 10%. The model is also able to predict the locking angle of 54.7∘ at higher pressures which is a distinct characteristic of McKibben actuators.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
A McKibben actuator/Pneumatic Artificial Muscle (PAM) is a soft actuator which has great potential in the field of bioinspired robotics. Miniaturized versions of PAMs or MPAMs of less than 1.5 mm diameter are ideal actuators for developing surgical devices due to their compliance and high power-to-weight ratio. Accurate mathematical model to represent the mechanics of PAM is an ongoing research. This paper develops a mathematical model which relates the input pressure to end-point deformation of a fabricated MPAM without external loading. The developed theoretical model is validated against experimental data for MPAM of lengths 60 and 70 mm. The model predicts the deformation of MPAM with standard error of less than 10%. The model is also able to predict the locking angle of 54.7∘ at higher pressures which is a distinct characteristic of McKibben actuators. |
| 5. | Menon, Midhun Sreekumar; Ravi, V C; Ghosal, Ashitava Trajectory planning and obstacle avoidance for hyper-redundant serial robots Journal Article Journal of Mechanisms and Robotics, 9 (4), pp. 041010:1-8, 2017. Abstract | BibTeX | Links: @article{Menon2017,
title = {Trajectory planning and obstacle avoidance for hyper-redundant serial robots},
author = {Midhun Sreekumar Menon and V. C. Ravi and Ashitava Ghosal},
url = {http://www.rbccps.org/wp-content/uploads/2018/12/jmr_009_04_041010.pdf},
doi = {10.1115/1.4036571},
year = {2017},
date = {2017-05-15},
journal = {Journal of Mechanisms and Robotics},
volume = {9},
number = {4},
pages = {041010:1-8},
abstract = {Hyper-redundant snakelike serial robots are of great interest due to their application in search and rescue during disaster relief in highly cluttered environments and recently in the field of medical robotics. A key feature of these robots is the presence of a large number of redundant actuated joints and the associated well-known challenge of motion planning. This problem is even more acute in the presence of obstacles. Obstacle avoidance for point bodies, nonredundant serial robots with a few links and joints, and wheeled mobile robots has been extensively studied, and several mature implementations are available. However, obstacle avoidance for hyper-redundant snakelike robots and other extended articulated bodies is less studied and is still evolving. This paper presents a novel optimization algorithm, derived using calculus of variation, for the motion planning of a hyper-redundant robot where the motion of one end (head) is an arbitrary desired path. The algorithm computes the motion of all the joints in the hyper-redundant robot in a way such that all its links avoid all obstacles present in the environment. The algorithm is purely geometric in nature, and it is shown that the motion in free space and in the vicinity of obstacles appears to be more natural. The paper presents the general theoretical development and numerical simulations results. It also presents validating results from experiments with a 12-degree-of-freedom (DOF) planar hyper-redundant robot moving in a known obstacle field.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Hyper-redundant snakelike serial robots are of great interest due to their application in search and rescue during disaster relief in highly cluttered environments and recently in the field of medical robotics. A key feature of these robots is the presence of a large number of redundant actuated joints and the associated well-known challenge of motion planning. This problem is even more acute in the presence of obstacles. Obstacle avoidance for point bodies, nonredundant serial robots with a few links and joints, and wheeled mobile robots has been extensively studied, and several mature implementations are available. However, obstacle avoidance for hyper-redundant snakelike robots and other extended articulated bodies is less studied and is still evolving. This paper presents a novel optimization algorithm, derived using calculus of variation, for the motion planning of a hyper-redundant robot where the motion of one end (head) is an arbitrary desired path. The algorithm computes the motion of all the joints in the hyper-redundant robot in a way such that all its links avoid all obstacles present in the environment. The algorithm is purely geometric in nature, and it is shown that the motion in free space and in the vicinity of obstacles appears to be more natural. The paper presents the general theoretical development and numerical simulations results. It also presents validating results from experiments with a 12-degree-of-freedom (DOF) planar hyper-redundant robot moving in a known obstacle field. |
