New England Manipulation Symposium (NEMS) 2026

University of Massachusetts Amherst, June 16-17, 2026

What is NEMS?

The New England Manipulation Symposium (NEMS) is an annual research symposium in the Northeast focused on robot manipulation research of all kinds, including mechanics, mechanisms, control, learning, perception, and planning. NEMS has two main goals. First, to provide an opportunity for members of the Northeast robot manipulation research community to meet and talk in person, share ideas, and become familiar with each other’s work. Second, to provide students and early-career researchers a platform from which to present their latest work. NEMS 2026 will be hosted by the University of Massachusetts Amherst.

This year at NEMS, we aim to encourage and highlight work that includes contributions to open-source and/or benchmarking. A subset of submissions will be highlighted throughout the symposium for serving as excellent examples of (a) leveraging existing open-source software or hardware, (b) contributing their work as open-source with sufficient documentation, (c) benchmarking the performance of their work by comparing to others, and/or (d) contributing a new benchmarking resource such as an evaluation protocol, artifact, or dataset.

Keynote Speakers

Day 1, 10:15

Heather Culbertson, University of Southern California

Learning to Feel: Touch, Perception, and the Future of Manipulation [slides]

Robot manipulation has advanced rapidly over the past decade, driven by breakthroughs in perception, machine learning, and large-scale data collection. Despite this progress, many manipulation systems remain overwhelmingly vision-centric. Physical interaction, however, is fundamentally about contact with the world, and touch provides information that vision alone cannot capture. Humans rely on touch to detect slip, estimate material properties, manipulate objects without looking, and understand interactions that are difficult to infer visually. Touch is central to how we interact with the physical world, yet it remains one of the most underutilized sources of information in robotic manipulation. In this talk, I'll explore what manipulation research can gain by taking touch seriously. I'll share work from my lab on tactile and acoustic sensing, data-driven models of physical interaction, and the science of how humans perceive touch. I'll also discuss how insights from human perception can help us design better sensors, build richer representations of contact, and create more capable manipulation systems. As robots move beyond structured environments and into the real world, touch will become increasingly important, not just as another sensor, but as a fundamental part of how intelligent systems understand and interact with their surroundings.

Day 1, 1:30

Rob Platt, Northeastern University

Model Matters: Leveraging Geometry in Robot Learning

Today, most robot learning models ignore the geometry of the environment, focusing instead on enlarging model and dataset size. Whereas this approach has been very successful in language and vision, geometry is more important in robotics than in those fields. How can we leverage geometric structural priors in robot learning? In this talk, I will summarize our recent work on this topic and make predictions about the future.

Day 2, 9:00

Ed Colgate, Northwestern University / HAND ERC

Strong Yet Backdrivable Robots

Dexterous manipulation requires combining high force output with passive backdrivability; capabilities that conventional geared actuation struggles to deliver. In this talk, I’ll introduce an electromechanical multiplexing architecture that routes power from a single drive shaft to multiple outputs and mechanically grounds them using capstan-amplified electroadhesive (EA) clutches in a load-transfer configuration. Wrapping thin-film EA clutches on cylindrical counter-surfaces provides exponential gain for EA braking force, while voltage pulse-width modulation yields sub-newton (<0.1N) force resolution and low reflected inertia from the drive shaft, enabling compliant interaction.

Keynotes: 45 mins (35 presentation + 10 Q&A) Sponsor and submitted talks: 15 mins (10 presentation + 3 Q&A + 2 wiggle room)

Day 1: Tuesday, June 16, 2026

9:00 Registration and breakfast
10:00 Introduction
10:15 🤖 Keynote: Learning to Feel: Touch, Perception, and the Future of Manipulation
Heather Culbertson, University of Southern California
11:00 🤖 Sponsor talk: SlipSense: Multimodal Tactile Learning for Low-Latency and Generalized Slip Detection
Analog Devices

Submitted talks:

11:15 Friction Compensation on Force Feedback Teleoperation for Improved Operator Comfort
Boon Yang Koh, Northeastern University
11:30 Loop closure grasping: Topological transformations enable strong, gentle, and versatile grasps
Kentaro Barhydt, Massachusetts Institute of Technology
11:45 Feature-Level Mixture-of-Experts for Robust Robotic Grasp Detection
Venkatesh Mullur, Worcester Polytechnic Institute
12:00 Deliberate Practice: Learning Robot Skills under a Budget
Shivam Vats, Brown University
12:15 Lunch break and poster session
1:30 🤖 Keynote: Model Matters: Leveraging Geometry in Robot Learning
Rob Platt, Northeastern University
2:15 🤖 Sponsor talk
Chewy Robotics

Submitted talks:

2:30 Manipulation-Guided Object Reconstruction for Cluttered Environments
Eric Kevin Wang, Massachusetts Institute of Technology
2:45 Toward Learning Manipulation Skills from Human Intervention Timing
Anjiabei Wang, Yale University
3:00 Learning Dynamic Rope Manipulation Using Task-Level Iterative Learning Control
Krishna Suresh, Carnegie Mellon University
3:15 When Life Gives You BC, Make Q-functions: Extracting Q-values from Behavior Cloning for On-Robot Reinforcement Learning
Thomas Weng, RAI Institute
3:30 Coffee break and poster session
4:00 🤖 Sponsor talk: From Classical Vision to End-to-End AI Policies for in Demand Industrial Robotic Tasks
Analog Devices

Submitted talks:

4:15 In-Hand Manipulation Planning for Grippers with Active Surfaces
Shambhuraj A. Mane, Worcester Polytechnic Institute
4:30 AURA: Asymptotically Optimal Uncertainty-Robust Replanning Algorithm for Kinodynamic Systems
Seyedali Golestaneh, Worcester Polytechnic Institute
4:45 Clutter Metrics to Evaluate Multi-Object Scenes for Benchmarking Robot Grasping and Manipulation
John Brann, University of Massachusetts Lowell
5:00 🤖 Reception and poster session
7:00 End of Day 1

Day 2: Wednesday, June 17, 2026

8:30 Registration and breakfast
9:00 🤖 Keynote: Strong Yet Backdrivable Robots
Ed Colgate, Northwestern University / HAND ERC
9:45 🤖 Sponsor talk
RAI Institute

Submitted talks:

10:00 DexWrist: A Robotic Wrist for Constrained and Dynamic Manipulation
Martin Peticco, Massachusetts Institute of Technology
10:15 Open-Source Manipulation Hardware
Vatsal Patel, Massachusetts Institute of Technology
10:30 Coffee break
11:00 Large and Fully Soft Suction Cup Arrays for Compliant and Preload-free Grasping
Qifan Yu, Massachusetts Institute of Technology
11:15 Tendon-driven soft grasper with asymmetric actuation through a jamming mechanism
Paola Romero, Massachusetts Institute of Technology
11:30 Toward Developing Guidelines for Modular Components in Robot Grasping and Manipulation Pipelines for Picking In Clutter
Huajing Zhao, University of Massachusetts Lowell
11:45 🤖 Tours
12:30 Lunch
1:30 🤖 Workshops
Workshop on Integration, Reproducibility, and Benchmarking of Robot Manipulation
Workshop on Soft Robotics for Manipulation
3:00 Coffee break
3:30 🤖 Workshops (continued)
5:00 End of Day 2

Workshops

In the afternoon of day 2 of NEMS 2026, we will hold two workshop sessions: Workshop on Integration, Reproducibility, and Benchmarking of Robot Manipulation, and Workshop on Soft Robotics for Manipulation. See below for more information.

🤖 Workshop on Integration, Reproducibility, and Benchmarking of Robot Manipulation

Benchmarking robot manipulation capabilities and comparing research solutions is either performed at the component level or holistically. Physical evaluations typically involve the latter which requires a full robot manipulation system. However, researchers often contribute a single novel software component – such as a grasp planner or perception module – while integrating multiple open-source products as part of the manipulation pipeline. The high dimensionality of a robot manipulation system makes it difficult to determine what factors contributed to the resulting performance and reproduce experiments. A lack of standards and guidelines on component structures, input/output formats (for pipeline integration), and test and evaluation procedures to ensure compatibility and usability places a significant burden on researchers. This workshop seeks to unite the robot manipulation research community towards the development of guidelines to improve integration of software and hardware, reproducibility of functionality and results, and benchmarking for side-by-side comparison.

This workshop will consist of informal panel discussions with NEMS contributors, questions/feedback from the audience, and live polling/crowdsourcing (using Mentimeter), split into two sessions: (1) developing and integrating open-source, and (2) benchmarking robot manipulation.
A set of NEMS contributions have been selected and grouped together to serve as jumping off points for each session, reviewing questions including:
- Session 1: What existing open-source software or hardware do you leverage in your own research? What contributions to open-source have you made, either your own work or contributions to existing work? What qualities of open-source have made it easier or harder to find, understand its applicability to your work, implement, evaluate, etc.?
- Session 2: Do you typically conduct benchmarking / performance evaluation to compare your work to others? What benchmarking assets have you contributed? E.g., evaluation protocol, artifact, object set, dataset, tools. Do you use existing benchmarking assets or did you have to create your own?
We will also review recommendations developed as part of the COMPARE Ecosystem project (https://robot-manipulation.org/), iterating with suggestions derived from the workshop, and developing new guidelines.

🤖 Workshop on Soft Robotics for Manipulation

While reproducibility is widely accepted as required for scientific progress, achieving reproducibility in soft robotics remains challenging due to the heavy dependence on custom, niche, novel or otherwise uncontrolled manufacturing and test methods. Variance at the manufacturing and material level leads to variance in end products, e.g., manipulators, and addressing these challenges while maintaining the flexibility and freedom necessary for research requires a community-level effort and agreement on what elements of manufacturing and testing procedures can, should and must be reported.

This workshop trials a hands-on and discussion-based approach to identifying these key parameters. Participants will engage in real soft robot manufacturing for the purposes of in-the-moment consideration of what parameters exist, potential sources of variance and reporting challenges. Following the hands-on activities, participants will engage in a discussion of findings.

Prior experience with soft robotics is encouraged but not strictly required.

Accepted Posters

Posters will be 24" x 36", vertical or horizontal.

A Trackball-Based Haptic Display for Characterizing Slip and Shear Rendering Performance: Philip LeShane, Ebenezer Yawlui, Yuri Gloumakov (University of Connecticut)
AURA: Asymptotically Optimal Uncertainty-Robust Replanning Algorithm for Kinodynamic Systems: Seyedali Golestaneh, Zhuoyun Zhong, Donghyung Lee, and Constantinos Chamzas (Worcester Polytechnic Institute)
Autonomous Underwater Navigation and Manipulation using Yolov8 and Sonar: Elliot Stark, Eric Huynh, Nolan Allen, Paul Robinette (University of Massachusetts Lowell)
Clutter Metrics to Evaluate Multi-Object Scenes for Benchmarking Robot Grasping and Manipulation: John Brann, Peter Gavriel, Brian Flynn, Adam Norton, Holly Yanco (University of Massachusetts Lowell)
Deliberate Practice: Learning Robot Skills under a Budget: Shivam Vats, Sudarshan Sunil Harithas, Mete Tuluhan Akbulut, Arvind Raghunathan, George Konidaris (Brown University, Mitsubishi Electric Research Laboratories)
DexWrist: A Robotic Wrist for Constrained and Dynamic Manipulation: Martin Peticco, Gabriella Ulloa, John Marangola, Nitish Dashora, Pulkit Agrawal (Massachusetts Institute of Technology)
Divide, Discover, Master: Skill Discovery and Reuse from Multi-Stage Tasks via Segmentation and Recall: Andrea Pierré, Brendan Hertel, Reza Azadeh (University of Massachusetts Lowell)
Dynamic Modeling of Vibration-Induced Friction Modulation for Controlled Sliding in Robotic In-Hand Manipulation: Samruddhi Naukudkar, Shambhuraj Mane, Berk Calli (Worcester Polytechnic Institute)
Enhanced Reactive Intelligence: The Roles of Predictive Vector Fields and Non-Convex Optimization in Real-time Planning and Control: Riddhiman Laha (Northeastern University)
Feature-Level Mixture-of-Experts for Robust Robotic Grasp Detection: Venkatesh Mullur, Vinayak Kapoor, Berk Calli (Worcester Polytechnic Institute)
Friction Compensation on Force Feedback Teleoperation for Improved Operator Comfort: Boon Yang Koh, Samuel Hibbard, Ryo Takei, John P. Whitney (Northeastern University)
From Classical Vision to End-to-End AI Policies for in Demand Industrial Robotic Tasks: Analog Devices
From Quasi-Static Point Contacts to Dynamic Patch Contacts using Non-Linear Complementarity Constraints: Haroon Hublikar, Riddhiman Laha, Seth Hutchinson (Northeastern University)
Hitting a Target with a Bullwhip: Simplified Representations for the Whip’s Dynamics: Aleksei Krotov, Dagmar Sternad (Northeastern University)
Hybrid Image-Based Visual Servoing for Continuum Robot Manipulation: Eda Guven, Berk Calli (Worcester Polytechnic Institute)
In-Hand Manipulation Planning for Grippers with Active Surfaces: Shambhuraj A. Mane, Berk Calli, Andrew S. Morgan (Worcester Polytechnic Institute and RAI Institute)
KaRMA: A Kinematic Metric for Fine Manipulation Ability in Robotic Hands: Martin Peticco, Pulkit Agrawal (Massachusetts Institute of Technology)
Koopman-based Convex Control Parametrization for Soft Robotic Limbs: A Data-driven Approach: Ran Jing, Taha Ondogan, Roberto Tron, Andrew P. Sabelhaus (Boston University)
Large and Fully Soft Suction Cup Arrays for Compliant and Preload-free Grasping: Qifan Yu, Kaitlyn Becker (Massachusetts Institute of Technology)
Learning-Based Programmable Locomotion Control for NiTi SMA-Actuated Bio-Inspired Soft Robotic Systems: Cinay Dilibal, Asheesh Lanba, Savas Dilibal (Dartmouth College, University of Southern Maine, Istanbul Gedik University)
Learning Dynamic Rope Manipulation Using Task-Level Iterative Learning Control: Krishna Suresh, Chris Atkeson (Carnegie Mellon University)
Loop closure grasping: Topological transformations enable strong, gentle, and versatile grasps: Kentaro Barhydt, O. Godson Osele, Sreela Kodali, Cosima du Pasquier, Chase M. Hartquist, H. Harry Asada, Allison M. Okamura (Massachusetts Institute of Technology, Stanford University, University of Florida)
Manipulation-Guided Object Reconstruction for Cluttered Environments: Eric Kevin Wang, Jungseok Hong, Pyae Sone Nyo Hmine, John J. Leonard (Massachusetts Institute of Technology)
Multi Agent Search for Mine Clearance with 3D Printed Legged Soft Hybrid Robots: Danelle Tuchman, Manvir Lamba, Markus P. Nemitz (Tufts University)
Path-Length Conservation with Adjustable Force Transmission in Voluntary-Open and Voluntary-Close Prosthetic Terminal Devices: Albert Gan, Miracle T. Uwakwe, Yuri Gloumakov (University of Connecticut)
PhaseHand: An Underactuated 1-DOF Grasp Changing Robot Hand: Miracle Uwakwe, Yuri Gloumakov (University of Connecticut)
Real-Time Forearm Ultrasound for Gloveless Capture of Hand Skill and Grasping Force Toward a Wearable Input Modality for Robot Learning from Demonstration: Daniel Alvarengaa, Bimbraw Keshavb, Haichong K. Zhangb (Worcester Polytechnic Institute)
Robot Learning from Failed and Successful Demonstrations with Elastic Maps: Brendan Hertel, Reza Azadeh (University of Massachusetts Lowell)
SlipSense: Multimodal Tactile Learning for Low-Latency and Generalized Slip Detection: Analog Devices
Streamlined Object Reconstruction and Synthetic Data Generation for Perception and Manipulation: Peter Gavriel, Graham Stelzer, Adam Norton (University of Massachusetts Lowell)
Tendon-driven soft grasper with asymmetric actuation through a jamming mechanism: Paola Romero, Qifan Yu, Kait Becker (Massachusetts Institute of Technology)
Textile overmolding for increased design space of strain limiting features in soft robotic graspers: Cat Arase, Charlotte Folinus, Qifan Yu, Kaitlyn Becker (Massachusetts Institute of Technology)
Toward Developing Guidelines for Modular Components in Robot Grasping and Manipulation Pipelines for Picking In Clutter: Huajing Zhao, Brian Flynn, Adam Norton, Holly Yanco (University of Massachusetts Lowell, University of Massachusetts Amherst)
Toward Learning Manipulation Skills from Human Intervention Timing: Anjiabei Wang, Shuangge Wang, Tesca Fitzgerald (Yale University)
Towards a unified inference framework for VLAs in the robotics field: Graham Stelzer (University of Massachusetts Lowell)
Trajectory Generation for Underactuated Soft Robot Manipulators using Discrete Elastic Rod Dynamics: Beibei Liu, Akua K. Dickson, Ran Jing, Andrew P. Sabelhaus (Boston University)
When Life Gives You BC, Make Q-functions: Extracting Q-values from Behavior Cloning for On-Robot Reinforcement Learning: Lakshita Dodeja, Ondrej Biza, Shivam Vats, Stephen Hart, Stefanie Tellex, Robin Walters, Karl Schmeckpeper, Thomas Weng (RAI Institute, Brown University, Northeastern University)

Location

NEMS 2026 will be held at UMass Amherst in the Computer Sciences Laboratories (CSL) Building, room E144, located at 130 Governors Drive, Amherst, MA 01003.

Wi-Fi Access

Instructions from: https://www.umass.edu/it/get-connected#umass

Campus guests can self-register via SMS to get wireless access.

Connect your device to the UMASS network and wait for the login page to appear. If a login page does not appear, open a web browser and go to umass.edu.
Under Guest Wireless Options, click Guest Self-Reg Form.
Follow the steps to enter your cellphone number to receive a PIN via text message.
Enter the PIN in the Access code box and accept the terms to get online.