Relevant Publications

Papers

• 

  S. Sidhik, M. Sridharan, D. Ruiken.
  "Towards an Online Framework for Changing-Contact Robot Manipulation Tasks".

  International Conference on Intelligent Robots and Systems (IROS), 27 September -- 1 October 2021, Prague, Czech Republic (Virtual).
  ABSTRACT: We describe a framework for changing-contact robot manipulation tasks, which require the robot to make and break contacts with objects and surfaces. The discontinuous interaction dynamics of such tasks make it difficult to construct and use a single dynamics model or control strategy for such tasks. For any target motion trajectory, our framework incrementally improves its prediction of when contacts will occur. This prediction and a model relating approach velocity to impact force modify the velocity profile of the motion sequence such that it is C-infinity smooth, and help achieve a desired force on impact. We implement this framework by building on our hybrid force-motion variable impedance controller for continuous-contact tasks. We evaluate our framework in the illustrative context of a robot manipulator performing sliding tasks involving multiple contact changes with surfaces of different properties.
  Submitted to: International Conference on Intelligent Robots and Systems (IROS), 27 September -- 1 October 2021, Prague, Czech Republic (Virtual).
  
  Cite this work: Copy BibTex; Copy plaintext;

  Paper
• 

  S. Sidhik, M. Sridharan, D. Ruiken.
  "Learning Hybrid Models for Variable Impedance Control of Changing-Contact Manipulation Tasks".

  Eighth Annual Conference on Advances in Cognitive Systems (ACS 2020), August 10-12, 2020, (Virtual).
  ABSTRACT: Many robot manipulation tasks comprise discrete action sequences characterized by continuous dynamics, while the transitions between these discrete dynamic modes are characterized by discontinuous dynamics. The individual modes may represent different types of contacts, surfaces, or other factors, and each mode and transition between the modes may require a different control strategy. This paper describes a piece-wise continuous, hybrid control framework for such manipulation tasks. The underlying representation enables the robot to automatically and efficiently detect the transitions between known modes, recognize new modes, and incrementally learn a dynamics model for variable impedance (i.e., stiffness) control in each mode, invariant to the direction of motion and the magnitude of applied forces. The framework is evaluated on a robot manipulator sliding an object along a surface to achieve a desired motion trajectory in the presence of changes in surface friction, applied force, or the type of contact between the object and the surface.
  Submitted to: Eighth Annual Conference on Advances in Cognitive Systems (ACS 2020), August 10-12, 2020, (Virtual).
  
  Cite this work: Copy BibTex; Copy plaintext;

  Paper
• 

  S. Sidhik*, M. Mathew*, M. Sridharan, J. Wyatt, M. Azad, A. Hayashi.
  "Online Learning of Feed-Forward Models for Task-Space Variable Impedance Control".

  International Conference on Humanoid Robots (Humanoids), October 15-17, 2019, Toronto, Canada.
  ABSTRACT: During the initial trials of a manipulation task, humans tend to keep their arms stiff in order to reduce the effects of any unforeseen disturbances. After a few repetitions, humans perform the task accurately with much lower stiffness. Research in human motor control indicates that this behavior is supported by learning and continuously revising internal models of the manipulation task. These internal models predict future states of the task, anticipate necessary control actions, and adapt impedance quickly to match task requirements. Drawing inspiration from these findings, we propose a framework for online learning of a time-independent forward model of a manipulation task from a small number of examples. The measured inaccuracies in the predictions of this model dynamically update the forward model and modify the impedance parameters of a feedback controller during task execution. Furthermore, our framework includes a hybrid force-motion controller that provides compliance in particular directions while adapting the impedance in other directions. These capabilities are evaluated on continuous contact tasks such as pulling non-linear springs, polishing a board, and stirring porridge.
  Submitted to: International Conference on Humanoid Robots (Humanoids)", October 15-17, 2019, Toronto, Canada.
  
  Cite this work: Copy BibTex; Other formats;

  Paper

Workshops

• 

  S. Sidhik, M. Sridharan, D. Ruiken
  "Towards a Framework for Changing-Contact Robot Manipulation".

  "Autonomous Robots and Multirobot Systems (ARMS) 2021 Workshop" at AAMAS 2021, May 3-7, 2021, London, UK (virtual).
  ABSTRACT: Many robot manipulation tasks require the robot to make and break contact with objects and surfaces. The dynamics of such changing-contact robot manipulation tasks are discontinuous when contact is made or broken, and continuous elsewhere. These discontinuities make it difficult to construct and use a single dynamics model or control strategy for any such task. We present a framework for smooth dynamics and control of such changing-contact manipulation tasks. For any given target motion trajectory, the framework incrementally improves its prediction of when contacts will occur. This prediction and a model relating approach velocity to impact force modify the velocity profile of the motion sequence such that it is C-infinity smooth, and help achieve a desired force on impact. We implement this framework by building on our hybrid force-motion variable impedance controller for continuous contact tasks. We experimentally evaluate our framework in the illustrative context of sliding tasks involving multiple contact changes with transitions between surfaces of different properties.
  Submitted to: The "Autonomous Robots and Multirobot Systems (ARMS) 2021 Workshop at the "20th International Conference on Autonomous Agents and Multiagent Systems", May 3-7, 2021, London, UK (virtual).

  Paper
• 

  S. Sidhik, M. Sridharan, D. Ruiken
  "Online Learning of Hybrid Models for Variable Impedance Control of a Changing-Contact Manipulation Task".

  Workshop on Interactive Learning at ICRA 2020, June 5, 2020, (Online).
  Many manipulation tasks comprise discrete action sequences characterized by continuous dynamics, with the transitions between these discrete dynamic modes characterized by discontinuous dynamics. The individual modes can represent different types of contacts, surfaces, or other factors, and different control strategies may be needed for each mode and the transitions between modes. This paper describes a piece-wise continuous, hybrid control framework that automatically detects transitions between modes and incrementally learns a model of the dynamics of each mode to support variable impedance control in that mode. The recognition of modes is invariant to the direction of motion and the magnitude of applied forces. Also, new modes are identified automatically and suitable models of the corresponding dynamics are learned. The framework is evaluated on a robot manipulator sliding an object on a surface along a desired motion trajectory in the presence of changes in surface friction, applied force, or the type of contact between the object and the surface. Experimental results indicate reliable and efficient recognition of modes, learning of dynamics models, and variable-impedance control during task execution.
  Submitted to: "Workshop on Interactive Learning" at the "International Conference on Robotics and Automation; ICRA 2020", June 5, 2020, (Online).

  Read Paper
• 

  S. Sidhik*, M. Mathew*, M. Sridharan, J. Wyatt, M. Azad, A. Hayashi.
  "Online Learning of Feed-Forward Models for Variable Impedance Control in Manipulation Tasks".

  Workshop on "Task-Informed Grasping: From Perception to Physical Interaction (TIG-II)" at the Robotics Science and Systems Conference (RSS), June 2019, Freiburg, Germany.
  ABSTRACT: While performing a new manipulation task, humans tend to be stiffer in the initial trials to improve task accuracy and counter any unforeseen disturbances. After a sufficient number of repetitions, humans are able to perform the task with lower stiffness without causing any significant reduction in task performance. Existing literature in human and animal motor control indicates that learned internal models of the manipulation task and the environment are used to predict the state in response to particular control actions, adapting stiffness on the fly to minimise energy usage. Inspired by these findings, we describe a framework that, in a significant departure from existing work, supports online learning of a time-independent forward model of any given manipulation task in the task space from a small number of samples. Errors in the predictions of the forward model are used to dynamically revise the model and adapt the impedance parameters of the feedback controller. Furthermore, a hybrid force-motion controller supports compliance in certain directions while adapting the stiffness in other directions. We evaluate our framework’s capabilities in the context of continuous-contact manipulation tasks with varying external forces.
  Submitted to: Workshop on "Task-Informed Grasping: From Perception to Physical Interaction (TIG-II)" at the "Robotics Science and Systems Conference (RSS)", June 22, 2019, Freiburg, Germany.
  
  

  Paper
• 

  S. Sidhik*, M. Mathew*, M. Sridharan, M. Azad, A. Hayashi
  "Online Learning of Forward Models for Variable Impedance Control in Manipulation Tasks".

  Workshop on "Emerging paradigms for robotic manipulation: from the lab to the productive world" at the Robotics Science and Systems Conference (RSS), June 22, 2019, Freiburg, Germany. WON BEST POSTER AWARD.
  Inspired by human motor control strategies, this work presents a general variable-impedance framework for robot manipulation that makes a significant departure from existing work. It comprises of a hybrid force-motion controller that learns a state-dependent forward model of any given manipulation task. The interplay between force and motion control enables the use of the errors in the predictions of the forward model to dynamically adapt the impedance parameters of the feedback controller. The time-independant formulation of variable impedance in the task-space of the robot makes it generalisable to different task environments.
  Submitted to: Workshop on "Emerging paradigms for robotic manipulation: from the lab to the productive world" at the "Robotics Science and Systems Conference (RSS)", June 22, 2019, Freiburg, Germany.

  Poster ("Best Poster Award" Winner)
• 

  S. Sidhik*, M. Mathew*, M. Sridharan, J. Wyatt, M. Azad, A. Hayashi.
  "Online Learning of Feed-Forward Models for Variable Impedance Control in Manipulation Tasks".

  UK Manipulation Worksphop, April 9-10, 2019, Leeds, UK.
  ABSTRACT: While performing a new manipulation task, it can be observed that humans tend to be stiffer in the initial trials to improve task accuracy and to counter any disturbances that may affect task performance. However, with more experience, humans create internal models of the task environment to predict environment state, which enables them to decrease the stiffness. This work aims to develop a similar variable impedance behaviour in robot manipulators by learning forward models of tasks. We put forward an approach to improve the feed-forward model online, and optimise the gain parameters of the robot's feedback controller using the model accuracy. Our framework also proposes a way of decoupling the ‘stiff’ and ‘compliant’ directions in a task using a hybrid force-motion controller.
  Submitted to: The "3rd UK Manipulation Worksphop, April 9-10, 2019, Leeds, UK.

  Poster

Open Source Software and Libraries

• 

  Franka ROS Interface

  A ROS/Python API for controlling and managing the Franka Emika Panda robot (real and simulated). Contains exposed controllers for the robot (joint position, velocity, torque), interfaces for the gripper, controller manager, coordinate frames interface, etc. Provides almost complete sim-to-real / real-to-sim transfer of code.
  Provides Python API and ROS utilites for controlling and managing the Franka Emika Panda robot (real and simulated). Contains exposed controllers for the robot (joint position, velocity, torque), interfaces for the gripper, controller manager, coordinate frames interface, etc. Also provides utilities to control the robot using MoveIt! and ROS Trajectory Action & ActionClient. This package also provides almost complete sim-to-real / real-to-sim transfer of code with the Panda Simulator package.
  
  Star
  Features
  

  • Low-level controllers (joint position, velocity, torque, impedance) available that can be controlled through ROS topics and Python API (including position control for gripper).
  • Real-time robot state (end-effector state, joint state, controller state, etc.) available through ROS topics and Python API.
  • Python API for managing controllers, coordinate frames, collision behaviour, controlling and monitoring the gripper.
  • Python API classes and utility functions to control the robot using MoveIt! and ROS Trajectory Action Service.
  • The panda_simulator package (which is Gazebo-based simulator for the robot) can also be controlled using this package (ROS and Python interface), providing almost complete sim-to-real transfer of code.

  Cite this work: Copy BibTex; Copy Plaintext;

  Code Repository API Documentation
• 

  Panda Simulator

  A Gazebo simulator for the Franka Emika Panda robot with ROS interface, providing exposed controllers and real-time robot state feedback similar to the real robot when using the franka_ros_interface package.
  A Gazebo simulator for the Franka Emika Panda robot with ROS interface, providing exposed controllers and real-time robot state feedback similar to the real robot when using the Franka ROS Interface package.
  
  Star
  Features
  

  • Low-level controllers (joint position, velocity, torque) available that can be controlled through ROS topics (including position control for gripper).
  • Real-time robot state (end-effector state, joint state, controller state, etc.) available through ROS topics.
  • The Franka ROS Interface package (which is a ROS interface for controlling the real Panda robot) can also be used with the panda_simulator, providing direct sim-to-real code transfer. The PandaRobot package which provides simplified python API for the robot can also be used for direct sim-to-real code transfer, as well as for performing real-time kinematics and dynamics computation.
  • Supports MoveIt planning and control for the robot arm and Franka Gripper.

  Cite this work: Copy BibTex; Copy Plaintext;

  Code Repository
• 

  Panda Robot

  A Python interface package built over the franka_ros_interface package for easier and managed controlling and handling of the Franka Emika Panda robot. Provides Kinematics and Dynamics computation for the robot. Also works directly with panda_simulator.
  A Python interface package built over the Franka ROS Interface package for easier and managed controlling and handling of the Franka Emika Panda robot. Provides Kinematics and Dynamics computation for the robot. Also works directly with panda_simulator.
  
  Star
  Features
  

  • Provides simple intuitive interface classes with methods to directly and easily control the robot.
  • Get real-time robot state, joint state, controller state, kinematics, dynamics, etc.
  • Provides Kinematics and Dynamics computation (using KDL library).
  • Integrated with gripper control.
  • Manage frames transformation and controller switching using simple utility functions.
  • Works directly on simulated robot when using panda_simulator providing direct sim-to-real and real-to-sim code transfer.

  Code Repository API Documentation
• 

  PyBullet Robot

  A generel Python interface class for robot simulations using PyBullet. Also provides Python API class to control and monitor the robot in the simulation. An implementation of subclass is available which uses the Franka Emika Panda robot as an example. An implementation of task-space hybrid force-motion controller is also available.
  A generel Python interface class for robot simulations using PyBullet. Also provides Python API class to control and monitor the robot in the simulation. The python API class provides functions to retrieve all physics information from the robot and environment, as well as provides interface methods to control the robot using the available controllers. The robot can be directly controlled using position, velocity, or torque control of joints. Example implementations of task-space control (Impedance control, Hybrid force-motion control) using joint torque control is also available.
  An implementation of subclass is available which uses the Franka Emika Panda robot as an example. The interface is written in the structure similar to the panda_robot ROS package used for controlling the real robot. The interface is written in the structure similar to the panda_robot ROS package used for controlling the real robot. This allows for direct transfer of code to the real robot (when using the panda_robot ROS package).

  Code Repository
• 

  pyLasaDataset

  Python library for using the LASA Handwriting Dataset. Primarily used for creating baseline 2D trajectories for robot manipulation.
  Python library for using the LASA Handwriting Dataset. Primarily used for creating baseline 2D trajectories for robot manipulation. Position and velocity trajectories that can be used for training and testing different learning and control framework for robot manipulators.

  Code Repository
• 

  Reflex TakkTile2 ROS Interface

  A Python interface package built over the reflex-ros-pkg package for easier controlling and handling the Reflex TakkTile2 hand.
  A Python interface package built over the reflex-ros-pkg package for easier controlling and handling the Reflex TakkTile2 hand.

  Code Repository

Most of my other works are publicly available from my Github page.

Other Projects

• 

  Graph-Based Simultaneous Localisation and Mapping (SLAM) on a 6-DoF Aerial Robot Using Monocular Camera (Masters Thesis)
  Read Thesis Code Repository
• 

  Visual Odometry and High Level Controller for an Aerial Robot (MSc Summer Mini-Project)
  Drone Controller Code Visual Odometry Code

  Read Report
• 

  Penny, the Security Guard Robot (MSc Intelligent Robotics Group Project)
  Visit Project Website