Speakers

Title: Pose Estimation and Galilean Space-Time in Robotics

Plenary Speaker: Robert Mahony (ANU, Canberra, Australia)

Abstract: 

Estimating the pose of a moving vehicle is a prerequisite for control of the vehicle or even just to interpret any data collected on that vehicle. These two talks discuss pose estimation from the perspective of exploiting the rigid-body and Galilean symmetry of a moving rigid-body. I will begin with the relatively simple case of rigid-body kinematics and develop a simple pose observer for this case. I use the Galilean symmetry to extend this to an inertial model that account for second order linear kinematics. Finally, I will discuss the case when the system is moving in a gravity field and the estimation is taken with respect to a rotating frame.

Title: A Unified Control Approach Design for Aerial Vehicles 

Plenary Speaker: Tarek Hamel (University de Cote d'Azur, France)

Abstract: 

Airplanes, helicopters, and other Vertical Take-Off and Landing (VTOL) vehicles, blimps, rockets, hydroplanes, marine ships, and submarines are generally under-actuated. They are essentially composed of a rigid body immersed in a fluid medium (air or water). They are commonly controlled via a propulsive thrust force directed along a body-fixed privileged axis and a torque vector with one (in the case of a marine ship) and either two or three (in the case of airships and submarines) independent components in charge of modifying the body’s orientation on a 2D-plane or in 3D-space. These vehicles are under-actuated because, apart from the direction associated with the thrust force, the other direction(s) of displacement is (are) not directly actuated. Yet, until some recent works, the structural similitude between all these vehicles has never been exploited to develop a general control framework. In this talk, I will give some insights to take a good step in this direction by proposing a unified control strategy that considers aerodynamic (resp. hydrodynamic) forces in the control design of a large class of aerial vehicles, including VTOL vehicles, rockets, airplanes, and convertible vehicles. 

Title: Aeromechanics for Control of Rotary Wing Aerial Vehicles 

Speaker: Abhishek (IIT Kanpur)

Abstract: 

Rotary wing aerial systems specially those based on helicopter configuration present a unique challenge due to its dynamics consisting of rotor coupled to a rigid body. A clear understanding of the aeromechanics is essential to model the dynamics for control design. The rotor blade in itself works as a second order system with high damping, this create significant delays in input-output realisation. This presentation would outline the development of coupled dynamics of rotor blades and fuselage using physics as well as system identification approaches. This approach has been used to solve problems of autonomous autorotation for helicopter and also for aggressive manoeuvring. Some of these examples would be discussed during the course of the presentation. 

Title: From Adaptive Observers to Output Feedback MPC 

Speaker: Shubhendu Bhasin (IIT Delhi)

Abstract: 

In this talk, I will provide a brief overview of adaptive observers. These observers use input-output information to estimate both the states and the unknown parameters of an uncertain dynamical system. The presentation will highlight the challenges encountered in establishing asymptotic stability of the state and parameter estimate error dynamics. Next, I will discuss our recent work on developing an adaptive observer framework that guarantees parameter convergence in less restrictive excitation situations. Finally, I will talk about how we leverage the adaptive observer structure to address the output-feedback MPC problem for uncertain systems. 

Title: Physical Human-Robot Interaction (PhyHRI) - From Factories to Homes

Speaker: Domenico Campolo (NTU Singapore)

Abstract: 

Although it is no longer surprising to hear from the news that some new AI-backed algorithm outperformed humans in some poem-writing contests or painting competitions, the difference between humans and robots is quite striking when it comes to physical manipulation: while modern algorithms can strategize better than humans, e.g. repeatedly outperforming any chess champion, when it comes to moving a single chess-piece or stacking up a few LEGO blocks, our best robots are no match for a 2-year old (who can actually build a LEGO tower while watching TV and munching a biscuit...).

It is clear from the argument above that we are definitely missing something when it comes to understanding physical tasks. In my opinion, one of the main reasons for this gap in understanding the sensorimotor aspects of physical tasks lies in the subtleties behind our sense of touch or haptics. 

In this talk, I will describe my two-decade old and still on-going journey in the realm of physical human-robot interaction. I will start by summarizing important lessons from the 80s, in the fields of both robotics and neuroscience which I found particularly inspiring for the design of simple but efficient control and learning algorithms.

Title: Optimal Control Design in Krotov Framework

Speaker: Tushar Jain (IIT Mandi)

Abstract: 

Every process, existing in nature or designed artificially, tends to operate in some optimal manner. Mathematically, the problem, then, is essentially optimizing the objective function subject to the process dynamics and possible state and input constraints. This talk will present the exposition and analysis of optimal control problems in the rather less-known Krotov framework. This framework provides sufficient conditions for global optimality and remains unexplored in the existing literature.     

In this framework, the optimal control problem is translated into an equivalent optimization problem via an ad-hoc selection of the so-called Krotov function. This latter equivalent optimization problem is then solved using the Krotov iterative method. In our work, we provide solution methodologies that avoid this iterative computation through a suitable selection of the above functional. Such a selection also results in different perspectives and insights with respect to the existing solution methodologies in the optimal control theory. In this talk, we will first revisit the standard finite-horizon linear quadratic control design problems in the Krotov framework. Then, the results are extended to the case of infinite-horizon linear quadratic problems. For this case, a linear matrix inequality is obtained as a necessary and sufficient condition for the equivalent optimization problem’s convexity. This explication nicely sheds light on the origin of the celebrated linear matrix inequality in the optimal control literature. We will also discuss the extension of similar ideas to a class of nonlinear optimal control problems. 

Next, the Krotov conditions are utilized to obtain results for suboptimal control design where the feedback gain matrix is norm-bounded. Contrariwise to the existing results where the upper bound on the cost is pre-specified, we propose a technique to compute the upper bound associated with the suboptimal control law. Suitable convex optimization problems are formulated to obtain an expression for the upper bound on the cost both in terms of the initial conditions and independent of the initial conditions. Finally, the exposition with respect to the existing results is detailed. Next, we will discuss the solution to the consensus protocol design problem in the Krotov framework. It presents a new alternative technique for computing the optimal consensus protocols for single integrators communicating over a complete topology. The problem is translated to solving a set of nonlinear algebraic equations in terms of the design parameters. Finally, the values of the parameters that ensure the closed-loop stability are chosen. 

In the case of other topologies, however, it is found that nonlinear algebraic equations do not admit any solution, thus necessitating the computation of suboptimal consensus protocols. Finally, this work demonstrates how the results derived for suboptimal linear quadratic regulator design can be used for suboptimal consensus design for a class of multi-agent systems communicating over undirected topologies. A detailed exposition of the existing results will also be discussed.

Title: Synergistic Control Systems for Human-Quadcopter Collaborative Tasks 

Speaker: Vineet Vashista (IIT Gandhinagar)

Abstract: 

This work explores the innovative domain of human-quadcopter interaction, focusing on the development of a control system that facilitates collaborative tasks between humans and aerial robots in outdoor environments. The research introduces a novel approach where a quadcopter and a human operator jointly transport a payload, leveraging a custom-built system for accurate command interpretation and state feedback. The control strategy is tailored to ensure natural and safe interaction, enhancing the smoothness of the collaborative effort. Experiments involving novice subjects validate the practicality of this approach, highlighting its potential for various applications. This research lays the groundwork for future advancements in human-robot collaboration, with implications for enhancing efficiency and safety in payload transportation tasks. 

Title: Control of Autonomous Systems against Complex Tasks: A Symbolic Approach 

Speaker: Pushpak Jagtap (IISc Bangalore) 

Abstract: 

Due to the increasing level of autonomy and rapid technological advancements in sensing, computing, and communication, nowadays, many real-world applications are expected to do complex tasks. These complex tasks can be formally represented using spatio-temporal logic specifications or (in)finite strings over automata. On the other hand, the modelling complexities in real-world applications, such as combination/interconnection of physical and cyber components, noisy dynamics, dependency on state history, lack of knowledge of the exact mathematical model, interconnection between subsystems, constraints posed by implementing hardware platforms, are increasing. These system- and task-level complexities make the formally correct synthesis of control algorithms challenging. Solving this problem is beyond the scope of conventional control theory and needs to utilize some concepts from computer science. In this talk, I will discuss how one can combine knowledge from different theories of control systems, computer science and AI/ML to synthesize formally verified controllers for complex control systems (i.e., containing the aforementioned modelling complexities) that ensure the satisfaction of complex logical specifications.

Title: Optimal and Stable Pose Estimators for Aerial Vehicles

Speaker: Sukumar Srikant (Systems and Control, IIT Bombay)

Abstract: 

We will discuss a discrete-time method for pose estimation using observers based on the Lagrange d’ Alembert principle. Since the pose observer is based on variational principles it satisfies weak optimality with respect to a suitable cost. We also carry out a Lyapunov analysis to prove convergence of estimates in the absence of noise. The aerial vehicle is assumed to have only camera and inertial sensors without a GPS. These results are therefore easily extendable to relative pose estimation as well. 

I will also show some simulator based results on relative pose estimation for non-cooperative satellites. An observer satellite equipped with stereo cameras is assumed to be observing a debris satellite. The aim is to predict the rotational motion of the debris satellite. The scenario is simulated in the Unreal Engine gaming simulator in order to carry out a software in loop simulation. A purely vision based estimation method is outlined as applicable to real-world non-cooperative satellites.

Title: Planetary Exploration with Autonomous Rover

Speaker: Rima Ghosh (ISRO Bangalore)

Abstract:

Rover plays an important role in planetary exploration missions. They serve as mobile laboratory traversing over challenging terrains to carry out in situ scientific experiments. In recent times, planetary rovers have made some significant discoveries both on Mars & Moon. Nasa rovers like Curiosity, Perseverance has provided proof of ancient lake beds on Mars. Similarly, Chandrayaan3 rover, Pragyan which landed on the moon on 23 Aug, 2023, has revealed important clues about the genesis of Moon. The alpha particle X-ray spectrometer (APXS) scientific payload abroad Pragyan, revealed evidence of a former magma ocean near the south pole. These findings will help in human colonization of Mars & Moon.

This talk will cover keys features of planetary rovers, like autonomy, navigation & guidance, which enables it to traverse over large distance and carry out experiments. I will also touch upon the various challenges faced while traversing over uneven terrain covered with boulders & craters. Path planning is a key feature which enables the rover to travel safely over rugged terrain without toppling. The rover takes images of the terrain & uses it to generate a map. Image processing & visual odometry plays an important part in path planning. The talk will also elaborate the difficulties faced during mobility due to slip. As a result of wheel soil interaction, rover tends to deviate from its intended path & hence this poses challenges in localization.

The talk will conclude with a brief description of Pragyan’s journey over the lunar surface for one lunar day. The unique surface terrain combined with sun movement, called for innovative methods to meet the mission objectives. Post landing, Pragyan Rover embarked on a 12-day journey of moonwalking near the south pole of the moon. This journey has helped scientist to get better in sight into the geological composition of lunar regolith.