Roni Permana Saputra

Research

Current projects

Autonomous navigation · Personal mobility

Autonomous personal mobility and mobility-on-demand

We develop SEATER (Single-passenger Electric Autonomous Transporter), a compact electric vehicle for first- and last-mile travel inside campuses, office complexes, and other managed areas, in indoor (I-SEATER) and outdoor (O-SEATER) variants. The aim is an autonomous mobility-on-demand service that is affordable enough to deploy in Indonesia.

The work covers the whole vehicle: platform design, docking and path tracking with nonlinear model predictive control, lane detection and lane keeping with learned perception, and a virtual-reality environment for driver testing. A new collaboration with Shanghai Jiao Tong University extends this towards semantic urban mapping and multimodal intelligent transport.

SEATER at the West Java Provincial Government complex, supporting the mobility of older people and people with disabilities.

Funding BRIN (2023, 2024), with Universitas Diponegoro; BRIN – LPDP – MOST China (2026), with Shanghai Jiao Tong University.

Docking with NMPCPath tracking with NMPCLane detectionLane keepingVR driving test

Service robots · Multi-robot systems

Autonomous delivery and service robots

DELL-E is our last-mile delivery robot for delivery-on-demand in managed areas, developed as a complete system: the robot, its autonomy software, and the connected services around it. The work has produced a registered patent and an industrial design for the robot, and a companion delivery application.

DELL-E, an autonomous delivery robot for last-mile and last-metre delivery in restricted-access areas such as housing estates, apartments, offices, and government facilities (BRINovasi, BRIN).

In parallel, we study mobile robots for flexible manufacturing, including Mecanum-wheeled platforms and safe navigation with model predictive control and control barrier functions, and are extending this to multi-robot systems for smart manufacturing.

Funding BRIN (2025), with Universitas Diponegoro; BRIN (2026).

Mecanum robot for flexible manufacturingSafe navigation with MPC and CBF

Field robotics · Multimodal sensing

Inspection and exploration robots

QUINS is a quadruped inspection system for industrial sites and high-risk zones, combining a legged platform with IoT connectivity and a multimodal sensor payload. Alongside the platform, we work on the perception a robot needs to understand such environments: semantic mapping, hazard mapping, and object detection with 360-degree cameras and 3D LiDAR.

Funding BRIN (2026), with Pertamina, Universitas Gadjah Mada, and Universitas Airlangga.

Semantic hazard mappingObject detection with 360° cameras and LiDARIndoor semantic mapping

Manipulation · Assistive robotics

Manipulation and assistive robots

On the manipulation side, we work on position, force, and admittance control of a six-axis arm, and on vision-guided grasping. On the assistive side, we develop on-device tactile intelligence for low-cost prosthetic hands using bio-based stretchable sensors, and contribute to a lower-limb exoskeleton for physiotherapy assistance.

Funding Toray Science and Technology Research Grant (2025); BRIN – LPDP (2024), with Telkom University Surabaya.

End-effector position controlAdmittance controlVision-guided grasping

Sensor fusion · State estimation

Sensor fusion and localisation

A ground vehicle has to know where it is, reliably, and with sensors that are affordable. We fuse wheel odometry, inertial measurements, consumer-grade GPS, and visual mapping through extended Kalman filtering, and evaluate the result on personal mobility vehicles and mobile robots, indoors and outdoors.

Results include fusion-based localisation for outdoor personal mobility vehicles, a multi-sensor localisation system for low-cost indoor navigation, fusion of an indoor positioning system with odometry, and a survey of GPS-aided sensor fusion. My current work extends this line towards learning-based sensor fusion for state estimation.

Outdoor fusion (IJIRA)Indoor localisation (IJIRA)EKF fusionSurveyIndoor positioning and odometry

Motion planning · Model predictive control

Motion planning and model predictive control

Model predictive control (MPC) is a common thread across our robots. We use it to plan and track motion within the robot's dynamics and constraints: docking and path tracking for personal mobility vehicles with nonlinear MPC, and safe navigation for Mecanum-wheeled robots by combining MPC with discrete-time control barrier functions.

Current doctoral work extends this to adaptive MPC for mobile robots on potholed tracks and to MPC with control barrier functions for safe navigation. It builds on HiDO-MPC, which I developed during my PhD for autonomous casualty extraction.

Docking with NMPCPath tracking with NMPCSafe navigation with MPC and CBFHiDO-MPC paper

Earlier work

Rescue robotics · Model predictive control

Autonomous casualty extraction

My doctoral research at Imperial College London developed ResQbot, a mobile rescue robot that approaches and extracts casualties. Hierarchical decomposed-objective model predictive control (HiDO-MPC) coordinates the motion objectives of the approach, and ResQbot 2.0 adds a stretcher-style design with an inflatable neck-securing device. A review of rescue robots for casualty extraction followed in 2025.

Funding Indonesia Endowment Fund for Education (LPDP) doctoral scholarship.

HiDO-MPC paperResQbot 2.0 paperReview paperHiDO-MPC code

Robot perception · Sim-to-real learning

Learning to detect casualties

Ground-projected point clouds provide a representation for detecting people lying on the ground. This work uses synthetic training data and augmentation to transfer a learned detector to real sensor observations.

IROS 2019 paperDeep_GPPC code