Mohit Mehndiratta

Mohit Mehndiratta

Doctoral Candidate, Nanyang Technological University, Singapore

Hello! I received a B.Tech. degree in aerospace engineering from Amity University, Uttar Pradesh, India, in 2012. In August 2015, I received an M.Sc. degree in aerospace engineering jointly offered by Technische Universität München (TUM), Germany and Nanyang Technological University (NTU), Singapore. I am currently a doctoral candidate (PhD. candidate) and pursuing his research at the School of Mechanical and Aerospace Engineering, NTU. My research primarily focuses on optimization-based learning control of unmanned aerial vehicles incorporating some machine learning techniques including Gaussian process regression and reinforcement learning.

download cv

Latest publications:

Education

Mohit Mehndiratta<

Research

Interests

  1. (Nonlinear) model predictive control

  2. Learning-based control

  3. Gaussian process for nonlinear regression

  4. Control and planning of ground/aerial robots

  5. Deep neural network-based modeling

Projects

Gaussian Process-based Learning Control of Aerial Robots for Precise Visualization of Geological Outcrops

Generating 3D field models, commonly known as virtual outcrops, is getting increasingly popular amongst geoscientific researchers. While an aerial robot is often employed for the mapping process that involves taking a series of overlapping images, the manual data collection gets challenging due to various reasons, namely: 1) piloted flight over a straight path for outcrops having curved/uneven surface results in distorted images due to consistent auto-focusing, 2) wind disturbances make it difficult even for skilled pilots to precisely maintain the desired overlap, and 3) hiring of a skilled pilot is expensive as the outcrop generation requires hours of data collection. To tackle these challenges, we propose to fully automate the data acquisition process in this work. In that vein, firstly, the designed navigation algorithm takes care of the trajectory generation for mapping. Whereas, the proposed learning-based control framework, i.e. position tracking nonlinear model predictive controller in conjunction with Gaussian process-based disturbance regression, facilitates a precise tracking of the generated path. Thanks to the long-short term memory feature of the designed GP model, the disturbance forces are accurately estimated even for increasing magnitude levels and time-periods.

Automated Tuning of Nonlinear Model Predictive Controller

The non-trivial weight tuning process is a major concern for safe realtime implementation of model predictive controller (MPC) especially over aerial robots. In most cases, users prefer the trial-and-error method to obtain these weighting parameters either on a real robot or in simulation. Whereas the former might be dangerous – especially with aerial robots –, the latter requires an accurate system model. To overcome these challenges, this work presents an active exploration-based tuning methodologythat significantly reduces the time and effort for MPC implementation and hence, can be fairly useful for unskilled MPC users. Besides, it utilizes a high fidelity deep neural network (DNN) model of the aerial robot that circumvents dangerous trials over real robots; while being real flight worthy at the same time, which facilitates their direct deployment over the real robots. Furthermore, to cater for several operational uncertainties – decreasing battery voltage, communication delays – that may not be captured within the DNN model, fine-tuning of the weight sets is also performed over the real robot. This essentially demonstrates the real flight tuning feasibility of the proposed algorithm.

Robust Tracking Control of Aerial Robots via a Simple Learning Strategy-based Feedback Linearization

To overcome the limitations of a traditional feedback linearrization control (FLC) method, this work presents a simple learning (SL) strategy to facilitate accurate tracking of aerial robots in unknown/uncertain environments. The SL strategy updates the controller gains and disturbance estimate within the feedback control law by minimizing a cost function which is defined based on the closed-loop error dynamics of the nominal system. The stability of the proposed approach is proven for a second-order uncertain nonlinear system. Also, it is illustrated that the SL strategy can find the global optimum point, and the controller gains and disturbance estimate converge to a finite value, thus resulting in a bounded control signal at the steady-state.

Constrained Instantaneous Learning-based Nonlinear Model Predictive Control for Aerial Package Delivery Robots

In accordance with the goal of utilizing aerial robots for daily operations in real application scenarios, an aerial robot must learn from its own experience and its interactions with the environment. This work presents an instantaneous learning-based control approach for the precise trajectory tracking of a 3D-printed aerial robot which can adapt itself to the changing working conditions. Considering the fact that model-based controllers suffer from lack of modeling, parameter variations and disturbances in their working environment, we observe that the presented learning-based control method has a compelling ability to significantly reduce the tracking error under aforementioned uncertainties throughout the operation. Three case scenarios are considered: payload mass variations on an aerial robot for a package delivery problem, ground effect when the aerial robot is hovering/flying close to the ground, and wind-gust disturbances encountered in the outdoor environment. In each case study, parameter variations are learned using nonlinear moving horizon estimation (NMHE) method, and the estimated parameters are fed to the nonlinear model predictive controller (NMPC).

Publications

Book Chapters

  1. Mohit Mehndiratta, Erkan Kayacan, Siddharth Patel, Erdal Kayacan and Girish Chowdhary, "Learning-based fast nonlinear model predictive control for custom-made 3D printed ground and aerial robots." In Handbook of Model Predictive Control, pp. 581-605. Birkhäuser, Cham, 2019.

Journal Papers

  1. Mohit Mehndiratta, Erkan Kayacan, Mahmut Reyhanoglu and Erdal Kayacan. "Robust Tracking Control of Aerial Robots via a Simple Learning Strategy-based Feedback Linearization." In IEEE Access. (Accepted).
  2. Mohit Mehndiratta and Erdal Kayacan. "A constrained instantaneous learning approach for aerial package delivery robots: onboard implementation and experimental results." In Autonomous Robots 43, no. 8 (2019): 2209-2228.
  3. Mohit Mehndiratta and Erdal Kayacan. "Receding horizon control of a 3 DOF helicopter using online estimation of aerodynamic parameters." In Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering (2017): 0954410017703414.

Conference Papers

  1. Mohit Mehndiratta and Erdal Kayacan. “Gaussian Process-based Learning Control of Aerial Robots for Precise Visualization of Geological Outcrops." In European Control Conference (ECC), 2020. (Accepted)
  2. Mohit Mehndiratta, Efe Camci and Erdal Kayacan. "Automated Tuning of Nonlinear Model Predictive Controller by Reinforcement Learning." In 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 3016-3021. IEEE, 2018.
  3. Mohit Mehndiratta and Erdal Kayacan. "Reconfigurable fault-tolerant NMPC for Y6 coaxial tricopter with complete loss of one rotor." In 2018 IEEE conference on control technology and applications (CCTA), pp. 774-780. IEEE, 2018.
  4. Mohit Mehndiratta, Erkan Kayacan and Erdal Kayacan. "A Simple Learning Strategy for Feedback Linearization Control of Aerial Package Delivery Robot." In 2018 IEEE Conference on Control Technology and Applications (CCTA), pp. 361-367. IEEE, 2018.
  5. Mohit Mehndiratta and Erdal Kayacan. "Online Learning-based Receding Horizon Control of Tilt-rotor Tricopter: A Cascade Implementation." In 2018 Annual American Control Conference (ACC), pp. 6378-6383. IEEE, 2018.
  6. Wilson Ying Jun Lee, Mohit Mehndiratta and Erdal Kayacan. "Fly without borders with additive manufacturing: a microscale tilt-rotor tricopter design." In Proceedings of the 3rd International Conference on Progress in Additive Manufacturing (Pro-AM 2018), pp 256–261, 2018.
  7. Ruddhi Gokhale, Yi Wan, Mohit Mehndiratta and Erdal Kayacan. "Fixed-wing vertical-takeoff-andlanding UAV with additive manufacturing: a dual-rotor version." In Proceedings of the 3rd International Conference on Progress in Additive Manufacturing (Pro-AM 2018), pp 250–255, 2018.
  8. Mohit Mehndiratta, Anna Prach and Erdal Kayacan. "Numerical Investigation of Gaussian filters with a combined type Bayesian filter for nonlinear state estimation." IFAC-PapersOnLine 49, no. 18 (2016): 446-453.
  9. Mohit Mehndiratta, Erdal Kayacan and Tufan Kumbasar. "Design and experimental validation of single input type-2 fuzzy PID controllers as applied to 3 DOF helicopter testbed." In 2016 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pp. 1584-1591. IEEE, 2016.

Teaching/Mentorship

Contact

mohit005@e.ntu.edu.sg skype: mohit.mehndiratta90 +65 83155236
  • Doctoral Student
  • School of Mechanical and Aerospace Engineering
  • Nanyang Technological University, Singapore