I am a researcher and software engineer in the field of robotics. My primary expertise lies in robotic manipulation and developing data-driven solutions for real-world robotic challanges. I worked on plenty of robotic projects where I utilised AI models for intelligent robot planning and control using visual, tactile, and propioception feedback.

I started working with robots in my M.Sc. in Mechatronics Engieering at a surgical robotic research center developing tele-operation controllers. I continued my journey by doing my PhD at Intelligent Manipulation Lab by conducting ground-braking research in tactile-based slip controllers in roboitc manipulation tasks. I am currently a postdoctoral research associate at Lincoln Institute for Agri-Food Technology working on autonomous navigation and 3D modelling of different types of crops. You can find demonstrations of my robotic projects bellow.

Robotic Projects

Bio-inspired Trajectory Modulation for Slip Control in Robot Manipulation Tasks

Nazari et al. Nature Machine Intelligence (2024)

In robotic pick-and-place tasks, preventing object slippage is critical. While grip force control is the traditional method, our research highlights trajectory modulation as a superior alternative in certain scenarios, especially when combined with predictive models. This approach offers a promising solution for enhancing robotic manipulation and slip control.

Deep Functional Predictive Control (deep-FPC): Robot Pushing 3-D Cluster using Tactile Prediction

Nazari et al. IROS (2023)

This project proposes a data-driven controller using tactile predictions and Functional Predictive Control (d-FPC) to manage Physical Robot Interaction (PRI) in complex pushing tasks. Tested on a robot pushing a strawberry stem, the d-FPC successfully controls 3D object movement, offering a promising solution for PRI in robotic manipulation.

Proactive slip control by learned slip model and trajectory adaptation

Nazari et al. CoRL (2022)

This paper introduces a novel slip control method for robotic manipulation. Traditional approaches rely on increasing grip force, which may not be feasible or could damage delicate objects. Instead, we propose a data-driven predictive controller that uses an action-conditioned slip predictor and a constrained optimizer to avoid slip during movements. Tested in real robot experiments, the method effectively controls slip, even for objects not encountered during training.

Autonomous Black Grass Detection and Removal in Wheat Fields using Unmanned Ground Vehicle

Nazari et al. JFR (2025)

A fully autonomous unmanned ground vehicle (UGV) equipped with GPS-based navigation is designed for efficient black grass removal in wheat fields. The system utilizes an RGB camera to detect black grass and identify crop rows, enabling precise targeting. A mechanical tool, with an adjustable horizontal degree of freedom, ensures that only weeds are removed while avoiding crop damage. Upon detecting black grass, the tool descends to remove it and retracts after covering a 5-meter distance. A safety feature, based on LiDAR data, halts the mission if a human presence is detected, resuming operations once the area is clear.

Crop Digital Twinning Using Multimodal 3D Pointcloud Data Integration

Sklar et al. IPPS8 (2025)

Digital twins of crops enhance monitoring and breeding by simulating growth and health. We tackled challenges in combining UAV’s global views with UGV’s detailed, plant-level data using a collaborative data collection pipeline. A calibration rack and data fusion aligned UAV’s GPS-based data with UGV’s optical frame, creating unified 3D models. Tested on miscanthus crops, this method produces detailed, efficient digital twins for precision agriculture.

Talks

• 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
  A talk on deep functional predictive conrtol for flexible objects manipulation.

• 6th Annual Conference on Robot Learning (CoRL) 2022
  A talk on proactive slip control using robot trajectory adaptation.

• 2021 Towards Autonomous Robotic Systems Conference (TAROS)
  A talk on multi-modal deep predictive models for tactile state estimation.