https://blog.pouceheure.fr/Recent content on Hugo POUSSEURHugoenMon, 10 Nov 2025 00:00:00 +0000https://blog.pouceheure.fr/projects/autosys-teleop/Mon, 10 Nov 2025 00:00:00 +0000https://blog.pouceheure.fr/projects/autosys-teleop/Demonstration Video Video 1 - Teleoperation Demo. (Train Demo) This Video 1: Teleoperation Demo. (Train Demo) illustrates the longitudinal control of the car via teleoperation. Example of speed saturation: 0:07 - 0:16 (limited by lateral acceleration) Example of braking: 1:20 (vehicle is stopped by teleop command) Context The vehicle autonomously plans a path and follows it. A teleoperator controls only the longitudinal speed using a joystick, similar to driving a train.https://blog.pouceheure.fr/projects/autosys-lidar-and-decision/Tue, 14 Oct 2025 00:00:00 +0000https://blog.pouceheure.fr/projects/autosys-lidar-and-decision/Demonstration Video 1 - Video demo of the LiDAR Detection pipeline Info: A video demonstration of the decision-making system is coming soon. Introduction In autonomous driving, the vehicle must be able to make a series of decisions, such as slowing down, stopping at an intersection, or waiting until the road is clear before proceeding. These decisions depend on the behavior of other road users. The system must therefore identify surrounding obstacles and determine which ones are relevant for decision-making.https://blog.pouceheure.fr/projects/autosys-traffic-light-detection/Thu, 15 May 2025 00:00:00 +0000https://blog.pouceheure.fr/projects/autosys-traffic-light-detection/Motivation As part of an autonomous driving project, I contributed to the development of a system enabling a vehicle to adapt its behavior to traffic lights. The vehicle adjusts its speed according to the current traffic light state. The project is divided into four main components: Matching: Identifying traffic light positions on the map Detection: Determining the current state of the traffic light Planning: Adjusting vehicle speed according to the detection result Control: Applying the computed velocity to the vehicle The detection pipeline is implemented in Python, while control is handled in C++ within a ROS2 environment.https://blog.pouceheure.fr/projects/dashboard/Thu, 01 May 2025 00:00:00 +0000https://blog.pouceheure.fr/projects/dashboard/Overview In robotics and system development, having a clear and flexible interface to visualize system states, sensor data, and real-time metrics is important. This project offers a lightweight dashboard framework designed for ROS 2, allowing quick creation of dashboards through simple configuration files. Video 1 - Demo video of the dashboard overlaid on video. The Video 1: Demo video of the dashboard overlaid on video. shows a demo of the dashboard running over a video stream.https://blog.pouceheure.fr/projects/rviz-virtual-camera/Thu, 01 May 2025 00:00:00 +0000https://blog.pouceheure.fr/projects/rviz-virtual-camera/Demonstration Video 1 - Demo: Example output from a virtual camera. The following demontration, Video 1: Demo: Example output from a virtual camera. , shows the video created from a virtual camera. The virtual camera is attached to a frame, fixed to vehicle. Motivation At first, I recorded the screen during experiments, but this was not a good solution because: process it’s more complex to setup, it uses too many resources, the video quality is low, I must keep the window and RViz view in the same place.https://blog.pouceheure.fr/projects/autosys-control/Fri, 20 Dec 2024 00:00:00 +0000https://blog.pouceheure.fr/projects/autosys-control/Demonstration Video 1 - Demo video: planning and control in action. Overview As part of a collaborative autonomous driving project, I contributed to the low-level control system responsible for translating navigation commands into actuator signals (steering and torque). This system is fully integrated into a ROS2-based autonomous stack and has been deployed and tested on a real electric vehicle with in-wheel motors. Fig. 1 - Interfaces between planning and control modules.https://blog.pouceheure.fr/projects/autosys-local-planning/Tue, 10 Dec 2024 00:00:00 +0000https://blog.pouceheure.fr/projects/autosys-local-planning/Demonstration Video 1 - Demo video, pedestrian avoidance using lidar detection. Video 2 - Demo video, slalom test. Video 3 - Demo video, curve test. Overview As part of a collaborative autonomous driving project, I contributed to the implementation of the core planning functionalities for autonomous driving. The navigation stack transforms a target position into a safe and efficient motion trajectory using environmental context and map information. The process is divided into three stages:https://blog.pouceheure.fr/projects/panel-bus/Fri, 11 Oct 2024 00:00:00 +0000https://blog.pouceheure.fr/projects/panel-bus/Motivation Bus schedules are difficult to follow because of uncertain traffic conditions. For this reason, I decided to use the available data to build a tool that allows quick visualization of the schedules for a given station. There is already a developed solution, but it requires several clicks before reaching the information. The idea is simple: make the information as raw and as fast to access as possible. The solution is therefore to display the buses of a station directly, accessible through a single link.https://blog.pouceheure.fr/articles/thesis/Thu, 28 Mar 2024 00:00:00 +0000https://blog.pouceheure.fr/articles/thesis/Abstract My thesis is based on shared navigation between the autonomous system and the human. In our research, we focus on command fusion. In our approach, both entities, the human and the autonomous system, simultaneously control the vehicle, and a module acquires their commands and performs the command fusion. This approach involves studying the intentions of both the human and the autonomous system to ensure the most appropriate fusion of their choices and to evaluate the decision-making of each entity.https://blog.pouceheure.fr/cv/Mon, 15 Jan 2024 00:00:00 +0000https://blog.pouceheure.fr/cv/Education 2025: Obtained the French National Qualification to Apply for Associate Professor (CNU – Section 27) 2020–2024: PhD in Robotics, Université de Technologie de Compiègne, France Thesis: Shared navigation in a cybernetic multi-agent autonomous system 2018–2019: Specialized Master’s in Robotics, CentraleSupélec (Metz, France) 2014–2019: Engineering Degree in Computing, ESILV (Paris, France) Ranked 1st in class during the 1st and 2nd years of the engineering cycle Academic & Professional Experience 2024–2025: Research Engineer (CNRS), Heudiasyc Laboratoryhttps://blog.pouceheure.fr/projects/intention-quantification/Sun, 10 Dec 2023 00:00:00 +0000https://blog.pouceheure.fr/projects/intention-quantification/Abstract As part of a collaborative research effort on shared control between an autonomous driving system and a human driver, I contributed to work focused on evaluating the quality of driving intentions. A driving intention is defined as a short sequence of planned control actions (speed, steering) over a time horizon: $$ I_t = \{(v,w)_{t+0\cdot\Delta t},(v,w)_{t+1\cdot\Delta t},...,(v,w)_{t+n\cdot\Delta t}\} \quad \text{Eq (1)} $$ For example, an intention can be expressed as a sequence of linear and angular velocities applied to the vehicle.https://blog.pouceheure.fr/projects/visual-control/Wed, 01 Feb 2023 00:00:00 +0000https://blog.pouceheure.fr/projects/visual-control/Abstract As part of a research project, I contributed to the development and testing of a Visual Servoing framework, a control strategy that links image feature variations to robot velocities. Applied to autonomous vehicles, this approach enables lane-keeping purely from camera input, without GPS or map dependency. It integrates lane detection using deep learning and transforms the output into features used in a control law that centers the vehicle in the lane.https://blog.pouceheure.fr/articles/cooperative-architecture/Sun, 01 Jan 2023 00:00:00 +0000https://blog.pouceheure.fr/articles/cooperative-architecture/Abstract A cooperative navigation architecture for the search and recognition of targets using aerial and ground vehicles is proposed in this paper. The architecture allows managing aerial and ground vehicles to autonomously perform different tasks independently or cooperatively. For our application, two main tasks are conceived: aerial monitoring of a surface to search for targets and target ground recognition. In the target aerial detection task, the aerial drone autonomously follows a trajectory computed to cover the entire surface to be monitored, searching for targets using vision algorithms.https://blog.pouceheure.fr/articles/general-multi-criteria/Sun, 01 Jan 2023 00:00:00 +0000https://blog.pouceheure.fr/articles/general-multi-criteria/Abstract Determining the admissibility and quality of driving intentions, generated by an automated intelligent vehicle and by a human driver, is a sine qua none task in a human-intelligent vehicle share navigation. Our paper proposes a generic method to quantify driving intentions. These intentions are defined by a sequence of velocities. This formulation is based on metrics already discussed in the literature. It proposes a way to use them to evaluate a state, making a judgment, and to extend this evaluation to a sequence of states.https://blog.pouceheure.fr/articles/machine-learning-controller/Sun, 01 Jan 2023 00:00:00 +0000https://blog.pouceheure.fr/articles/machine-learning-controller/Abstract Researchers have been attempting to make the car drive autonomously. The environment perception together with safe guidance and control is an important task and are one of the big challenges when developing this kind of system. Geometrical or physical based models, machine learning based models and those based on a mixture of both models, are the three types of navigation methods used to resolve this problem. The last method takes advantage of the learning capability of machine learning models and uses the safeness of geometric models in order to better perform the navigation task.https://blog.pouceheure.fr/articles/camera-based-control/Sun, 01 Jan 2023 00:00:00 +0000https://blog.pouceheure.fr/articles/camera-based-control/Abstract By utilizing the camera installed in the front of a vehicle, this paper proposes on-board camera-based driving force control (CDFC) methods for autonomous electric vehicles driven by in-wheel motors. The image processing algorithm can detect the change in road surface conditions quickly and accurately. This enables the CDFC to update the slip-ratio limiter in real-time. Test results show that the proposed methods can improve traction control performance and reduce inverter input energy.https://blog.pouceheure.fr/projects/friction-detection/Thu, 10 Nov 2022 00:00:00 +0000https://blog.pouceheure.fr/projects/friction-detection/Mission Context As part of a European research project, I contributed to a two-month research mission conducted at the University of Tokyo, focusing on road surface friction estimation using only a forward-facing camera. This work led to the publication of a paper, ProposalTakumi 2023 , at IEEE AIM 2023, in collaboration with the University of Tokyo and UTC. The system uses image segmentation, confidence modeling, geometric projection, and accumulation into a global surface grid map.https://blog.pouceheure.fr/projects/multi-robots/Wed, 10 Aug 2022 00:00:00 +0000https://blog.pouceheure.fr/projects/multi-robots/Context As part of a research internship project, a modular ROS architecture was designed to enable cooperative navigation between heterogeneous autonomous vehicles: drones (UAVs) and ground robots (UGVs). The contribution focused on supporting the development and implementation of this system, which aimed to be scalable, with dynamic task allocation and coordinated execution across different hardware platforms. The architecture uses a unified communication protocol, distributed ROS nodes, and a namespace layout allowing multiple robots to be managed from a central ROS master.https://blog.pouceheure.fr/projects/human-prediction/Thu, 10 Mar 2022 00:00:00 +0000https://blog.pouceheure.fr/projects/human-prediction/Abstract A multimodal deep learning model was developed to predict human driving behavior over a short time horizon. The training dataset was recorded specifically for this project. Video 1 - Video demo, human driving behavior prediction. More details about this work are provided in PredictionPousseur 2022 . Introduction Driving intention is represented as a sequence of vehicle states. Let $I$ be: $$ I = \{x_0, x_1, ..., x_n\} \quad \text{Eq (1)} $$ Here, $(x_i) = (v_i, w_i)$ denotes the state at time $i$, where $v_i$ is the linear velocity and $w_i$ is the angular velocity.https://blog.pouceheure.fr/projects/lane-detection/Mon, 10 Jan 2022 00:00:00 +0000https://blog.pouceheure.fr/projects/lane-detection/Abstract Lane detection is an essential function for autonomous vehicles. While GPS and HD maps offer accurate localization under optimal conditions, vision-based detection provides an important backup, especially in areas with weak GNSS signals or outdated maps. This system implements a deep-learning lane detection pipeline using a convolutional autoencoder, designed to identify and segment multiple lane boundaries in various road conditions using only front-facing camera images. The approach integrates image preprocessing, deep-learning inference, optional temporal tracking, and curve fitting to generate stable and clean lane boundaries.https://blog.pouceheure.fr/articles/context-awareness/Sat, 01 Jan 2022 00:00:00 +0000https://blog.pouceheure.fr/articles/context-awareness/Abstract Many studies faced the problem of vehicle autonomous navigation in different fields, but nowadays just a few of them uses all the implicit information coming from the context in which such navigation is occurring. This results in a huge potential information loss that prevents us from adapting the vehicle’s behavior to each different situation it may be in. In a previous work, we defined a method to model the static context of navigation using ontologies and take it into account in the command law when performing a local navigation task.https://blog.pouceheure.fr/articles/gradient-descent/Sat, 01 Jan 2022 00:00:00 +0000https://blog.pouceheure.fr/articles/gradient-descent/Abstract Avoiding obstacles is a key feature in a vehicle autonomous navigation methodology. The Dynamic Window Approach (DWA), proposed several decades ago, has emerged as a responsive navigation methodology suitable for reactive obstacle avoidance. In the initial approach of the DWA, the optimization of an objective function is realized with an exhaustive computation, which can be costly in computational time. This is not useful in a real-time scenario where an autonomous vehicle needs to avoid obstacles in urban or road velocity conditions.https://blog.pouceheure.fr/articles/motion-control/Sat, 01 Jan 2022 00:00:00 +0000https://blog.pouceheure.fr/articles/motion-control/Abstract In this paper, a navigation control for a platoon composed of a ground and an aerial vehicle is proposed. The aim of this work is to define a platooning system that exploits the advantages of both types of robots: while the aerial drone is faster and is not affected by the terrain, the ground vehicle has more autonomy and can carry heavier payloads. The proposed control system allows the aerial vehicle to follow the ground vehicle, maintaining a desired relative position, and to assist in navigation by providing information about the environment.https://blog.pouceheure.fr/articles/human-driving-prediction/Sat, 01 Jan 2022 00:00:00 +0000https://blog.pouceheure.fr/articles/human-driving-prediction/Abstract Predicting the intentions of the human and the machine on a near future is required to the human-machine shared control of automated intelligent vehicles. The autonomous system is able to inform about its future intentions, however it is not possible for the human to provide this information, it is necessary then to predict it. This paper proposes a deep learning methodology to predict human navigation intentions in a time horizon of a few seconds, using a recurrent neural network (RNN) architecture based on the Long Short-Term Memory (LSTM) architecture.https://blog.pouceheure.fr/projects/dwa-optimization/Fri, 10 Dec 2021 00:00:00 +0000https://blog.pouceheure.fr/projects/dwa-optimization/Overview The Dynamic Window Approach (DWA) is a reactive motion planning method used in mobile robotics. At each control cycle, it samples admissible velocity pairs $(v, \omega)$ and evaluates them using an objective function, selecting the pair with the highest score. The original method presents two main issues: Discrete sampling, which limits precision High computational cost in evaluation loops These limitations can be addressed by defining a convex, differentiable objective function and applying gradient descent for continuous optimization.https://blog.pouceheure.fr/articles/context-modelling/Fri, 01 Jan 2021 00:00:00 +0000https://blog.pouceheure.fr/articles/context-modelling/Abstract This paper faces the problem of intelligent vehicles in interaction with their occupants and the environment, by modelling the semantic context associated with navigation. By semantically modelling context, an intelligent vehicle can not only drive itself safely but also reason about the situation and act accordingly. To achieve this, it is necessary to first define the Context of Navigation and then establish inference rules to enrich the vehicle’s understanding of the situation.https://blog.pouceheure.fr/articles/isolated-area-consumption/Fri, 01 Jan 2021 00:00:00 +0000https://blog.pouceheure.fr/articles/isolated-area-consumption/Abstract Forecasting consumption in isolated areas represents a challenging problem typically resolved using deep learning or large mathematical models with various dimensions. These models require expertise in metering and algorithms, and the equipment needs to be frequently maintained. In the context of the MAESHA H2020 project, most of the consumers and producers are isolated. Forecasting becomes more difficult due to the lack of external data and the significant impact of human behaviors on these small systems.https://blog.pouceheure.fr/articles/shared-decision/Fri, 01 Jan 2021 00:00:00 +0000https://blog.pouceheure.fr/articles/shared-decision/Abstract In this paper, a non-haptic shared control is applied to navigate an intelligent vehicle based on two inputs: one by a human driver and the other by an autonomous system. The proposed shared decision-making model aims to improve the safety and efficiency of autonomous vehicle navigation by integrating human intuition and machine precision. Experimental results demonstrate the effectiveness of this approach in various driving scenarios. :contentReference[oaicite:0]{index=0}https://blog.pouceheure.fr/projects/human-legs-detection/Thu, 10 Dec 2020 00:00:00 +0000https://blog.pouceheure.fr/projects/human-legs-detection/Overview Source Code In addition to the main GitHub project, the following submodules are available: Dataset of scanned legs (with labels): https://github.com/PouceHeure/dataset_lidar2D_legs Labeling GUI tool: https://github.com/PouceHeure/lidar_tool_label Radar interface: https://github.com/PouceHeure/ros_pygame_radar_2D Context This project detects human legs from 2D LiDAR scans using a Recurrent Neural Network (RNN) with LSTM cells. The LSTM processes each scan as a spatial sequence ordered by angle $ \theta $, not as a time series. The model learns local shape patterns in polar space that are typical of legs.https://blog.pouceheure.fr/projects/light-mas/Mon, 07 Sep 2020 00:00:00 +0000https://blog.pouceheure.fr/projects/light-mas/Overview py_light_mas is a lightweight Python framework for multi-agent systems (MAS). It provides simple abstractions for Agents, an Environnemnt, a Simulation loop, and an optional Network for inter-agent communication. Architecture The architecture can be illustrated in two diagrams that describe the same system: Fig. 1 - Relations between Simulation, Environnemnt, Agents, and Network. The Fig. 1 highlights how the Simulation connects the Environnemnt, Agents, and optional Network. Fig. 2 - Event sequence across a simulation tick with message passing.https://blog.pouceheure.fr/projects/pouco2000/Mon, 15 Jun 2020 00:00:00 +0000https://blog.pouceheure.fr/projects/pouco2000/Overview Pouco2000 is a C++-based project providing a modular physical control panel for interacting with ROS-based robots. The system includes: An Arduino library for defining hardware inputs and outputs ROS packages for serial communication, message extraction, and parameter introspection Utilities for visual monitoring and debugging in real-time Video 1 - Video demo, hardware and software test. Motivation Using only terminal commands or software interfaces to control and debug robots can be slow, especially in field environments.https://blog.pouceheure.fr/articles/jade-microgrids/Wed, 01 Jan 2020 00:00:00 +0000https://blog.pouceheure.fr/articles/jade-microgrids/Abstract This work presents a JADE-based model for optimizing microgrid energy distribution. The proposed approach utilizes a multi-agent system framework to manage and control various distributed energy resources within a microgrid. By implementing intelligent agent behaviors, the system aims to enhance energy efficiency, reliability, and scalability in smart grid applications. Simulation results demonstrate the effectiveness of the JADE-based model in handling dynamic energy demands and supply conditions, contributing to the advancement of intelligent energy management systems.https://blog.pouceheure.fr/projects/drone-control/Fri, 01 Feb 2019 00:00:00 +0000https://blog.pouceheure.fr/projects/drone-control/Overview This project was part of a final-year master’s program and involved developing a drone capable of autonomously flying over a predefined area. The work was split into two main parts: path planning and control. My responsibility was to control a Parrot Bebop drone using ROS, leveraging an existing ROS package that communicates with Parrot’s SDK. Video 1 - Drone control demonstration. Additional Features Controller Watchdog: Added a manual override system allowing the operator to take full control of the drone if the autonomous system fails.https://blog.pouceheure.fr/articles/multi-agent-smart-houses/Mon, 01 Jan 2018 00:00:00 +0000https://blog.pouceheure.fr/articles/multi-agent-smart-houses/Abstract Most demand-side management programs focus on the interactions between an aggregator and its users. Moreover, renewable energy production is irregular; increasing their number implies predicting consumption and managing energy storage or discharge in real time. Therefore, organizing the consumption patterns of every device connected to the grid is essential to optimize the global consumption. Studying the smart grid through modeling and simulation provides valuable results that cannot be obtained in the real world due to time and cost constraints.https://blog.pouceheure.fr/projects/lovo/Mon, 07 Aug 2017 00:00:00 +0000https://blog.pouceheure.fr/projects/lovo/Overview This project is a bicycle theft detection system based on the Sigfox network. It uses an Arduino board combined with an inertial measurement unit (IMU). When abnormal motion is detected, such as acceleration above a set threshold, the device sends data via Sigfox. The information is then forwarded to Firebase, which triggers a notification to the user’s mobile application. Video 1 - Video demo: the system sends a notification when motion is detected.