Gianni A. Di Caro
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Gianni A. Di Caro has a total of 125 co-authors. Here are the top co-authors by number of shared publications:
Luca M. Gambardella
Dalle Molle Institute for Artificial Intelligence (IDSIA), Lugano, Switzerland
36 shared publications- Alessandro Giusti
Dalle Molle Institute for Artificial Intelligence (IDSIA), SUPSI, Lugano, Switzerland
19 shared publications Frederick Ducatelle
USI-SUPSI, IDSIA, Manno, 6928, SWITZERLAND
18 shared publicationsJawad Nagi
Dalle Molle Instituite for Artificial Intelligence (IDSIA), Lugano, Switzerland
13 shared publicationsLuca Maria Gambardella
Dalle Molle Institute for Artificial Intelligence (IDSIA USI-SUPSI), Lugano, Switzerland
13 shared publicationsEduardo Feo Flushing
12 shared publications
Latest Publications
- Open Access
- 11 November 2026
- Published by Springer Nature in Journal of Intelligent & Robotic Systems
Deep Learning-Based Fast Inference of Team Performance in Heterogeneous Multi-Robot Systems
The effectiveness of multi-robot teams depends on various factors, including the size of the team, the capabilities of each individual robot, the synergies between them, and the coordination strategies employed. However, current studies on multi-robot task allocation often assume pre-established teams, which results in a focus on optimizing resources rather than designing teams from a diverse range of robots or selecting the most suitable team from a set of potential configurations. Our research focuses on predicting a team’s performance in a given mission based on its composition. This task has significant implications for selecting or designing multi-robot teams, as it helps to optimize performance while minimizing costs. To this end, we propose a novel fast inference method for estimating team performance using supervised learning in multi-robot task allocation situations. We considered a general multi-robot task allocation issue for spatially distributed missions. For evaluation purposes, we used a large dataset of varied problem instances, two different solvers, and five different team performance metrics derived from the solutions computed by these solvers. The results demonstrate that long short-term memory recurrent neural networks, aided by convolutional layers, can accurately predict team performance metrics such as makespans and intra-team communication metrics derived from the solutions obtained by the two solvers. We also highlight the efficacy of transfer learning in fine-tuning models from a simpler sub-optimal solver to a more complex, optimal solver, demonstrating improved accuracy with reduced training set size and computational burden.
- 11 November 2026
- Published by Springer Nature in Journal of Intelligent & Robotic Systems
Open AccessDeep Learning-Based Fast Inference of Team Performance in Heterogeneous Multi-Robot Systems
The effectiveness of multi-robot teams depends on various factors, including the size of the team, the capabilities of each individual robot, the synergies between them, and the coordination strategies employed. However, current studies on multi-robot task allocation often assume pre-established teams, which results in a focus on optimizing resources rather than designing teams from a diverse range of robots or selecting the most suitable team from a set of potential configurations. Our research focuses on predicting a team’s performance in a given mission based on its composition. This task has significant implications for selecting or designing multi-robot teams, as it helps to optimize performance while minimizing costs. To this end, we propose a novel fast inference method for estimating team performance using supervised learning in multi-robot task allocation situations. We considered a general multi-robot task allocation issue for spatially distributed missions. For evaluation purposes, we used a large dataset of varied problem instances, two different solvers, and five different team performance metrics derived from the solutions computed by these solvers. The results demonstrate that long short-term memory recurrent neural networks, aided by convolutional layers, can accurately predict team performance metrics such as makespans and intra-team communication metrics derived from the solutions obtained by the two solvers. We also highlight the efficacy of transfer learning in fine-tuning models from a simpler sub-optimal solver to a more complex, optimal solver, demonstrating improved accuracy with reduced training set size and computational burden.
- 14 September 2026
- Published by Springer Nature
- p. 631-643
Hybrid Analytical–Learned Representations for Preserving Machine-Readable Patterns in Gaussian Splatting
- 14 September 2026
- Published by Springer Nature
- p. 631-643
Hybrid Analytical–Learned Representations for Preserving Machine-Readable Patterns in Gaussian Splatting
- Open Access
GreenhouseSplat: A Dataset of Photorealistic Greenhouse Simulations for Mobile Robotics
Simulating greenhouse environments is critical for developing and evaluating robotic systems for agriculture, yet existing approaches rely on simplistic or synthetic assets that limit simulation-to-real transfer. Recent advances in radiance field methods, such as Gaussian splatting, enable photorealistic reconstruction but have so far been restricted to individual plants or controlled laboratory conditions. In this work, we introduce GreenhouseSplat, a framework and dataset for generating photorealistic greenhouse assets directly from inexpensive RGB images. The resulting assets are integrated into a ROS-based simulation with support for camera and LiDAR rendering, enabling tasks such as localization with fiducial markers. We provide a dataset of 82 cucumber plants across multiple row configurations and demonstrate its utility for robotics evaluation. GreenhouseSplat represents the first step toward greenhouse-scale radiance-field simulation and offers a foundation for future research in agricultural robotics.
Open AccessGreenhouseSplat: A Dataset of Photorealistic Greenhouse Simulations for Mobile Robotics
Simulating greenhouse environments is critical for developing and evaluating robotic systems for agriculture, yet existing approaches rely on simplistic or synthetic assets that limit simulation-to-real transfer. Recent advances in radiance field methods, such as Gaussian splatting, enable photorealistic reconstruction but have so far been restricted to individual plants or controlled laboratory conditions. In this work, we introduce GreenhouseSplat, a framework and dataset for generating photorealistic greenhouse assets directly from inexpensive RGB images. The resulting assets are integrated into a ROS-based simulation with support for camera and LiDAR rendering, enabling tasks such as localization with fiducial markers. We provide a dataset of 82 cucumber plants across multiple row configurations and demonstrate its utility for robotics evaluation. GreenhouseSplat represents the first step toward greenhouse-scale radiance-field simulation and offers a foundation for future research in agricultural robotics.
- Open Access
- 22 July 2022
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 17 (7) , e0271751
A clustering metaheuristic for large orienteering problems
The Orienteering Problem is a routing problem that commonly appears in mapping, transportation, distribution, and scheduling scenarios. The Orienteering Problem is a challenging NP-hard problem, such that obtaining optimal or near optimal solutions usually requires a significant amount of computation for large and even moderately large instances. As a result, existing algorithms cannot effectively be utilized for solving large Orienteering Problems in an online setting, which is often required in real-world applications where a plan has to be iteratively computed. For instance, a planner might need to adapt to changes in the scenario or in the environment (e.g., this is common in goods delivery, as well as in mobile robotic scenarios for coverage, monitoring, and surveillance). Motivated by these limitations, we propose a multi-stage clustering-based metaheuristic for tackling large Orienteering Problems in an effective, strategically controlled amount of computation time. The metaheuristic starts by decomposing the problem into smaller, independent sub-problems that are efficiently solved using an algorithm of choice, sequentially or in parallel. The obtained solutions are then merged to form a solution for the original problem, and then further optimized and processed to ensure feasibility. The metaheuristic aims to dramatically improve the computation time of a given Orienteering Problem algorithm without a significant decrease in the solution quality of that algorithm, especially for large Orienteering Problems. We have validated the effectiveness of the proposed metaheuristic design through a set of computational experiments. In particular, using a state-of-the-art heuristic and an exact algorithm, we have shown that it is significantly beneficial to use the Orienteering Problem algorithm plugged into our metaheuristic, as opposed to using it as a standalone algorithm. This was found to be especially true for large problems. As a result, large instances of Orienteering Problems can be effectively tackled within reasonable time bounds even for online application scenarios.
- 22 July 2022
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 17 (7) , e0271751
Open AccessA clustering metaheuristic for large orienteering problems
The Orienteering Problem is a routing problem that commonly appears in mapping, transportation, distribution, and scheduling scenarios. The Orienteering Problem is a challenging NP-hard problem, such that obtaining optimal or near optimal solutions usually requires a significant amount of computation for large and even moderately large instances. As a result, existing algorithms cannot effectively be utilized for solving large Orienteering Problems in an online setting, which is often required in real-world applications where a plan has to be iteratively computed. For instance, a planner might need to adapt to changes in the scenario or in the environment (e.g., this is common in goods delivery, as well as in mobile robotic scenarios for coverage, monitoring, and surveillance). Motivated by these limitations, we propose a multi-stage clustering-based metaheuristic for tackling large Orienteering Problems in an effective, strategically controlled amount of computation time. The metaheuristic starts by decomposing the problem into smaller, independent sub-problems that are efficiently solved using an algorithm of choice, sequentially or in parallel. The obtained solutions are then merged to form a solution for the original problem, and then further optimized and processed to ensure feasibility. The metaheuristic aims to dramatically improve the computation time of a given Orienteering Problem algorithm without a significant decrease in the solution quality of that algorithm, especially for large Orienteering Problems. We have validated the effectiveness of the proposed metaheuristic design through a set of computational experiments. In particular, using a state-of-the-art heuristic and an exact algorithm, we have shown that it is significantly beneficial to use the Orienteering Problem algorithm plugged into our metaheuristic, as opposed to using it as a standalone algorithm. This was found to be especially true for large problems. As a result, large instances of Orienteering Problems can be effectively tackled within reasonable time bounds even for online application scenarios.
- 1 January 2022
- Published by Springer Nature
- p. 335-348
Map Learning via Adaptive Region-Based Sampling in Multi-robot Systems
- 1 January 2022
- Published by Springer Nature
- p. 335-348
Map Learning via Adaptive Region-Based Sampling in Multi-robot Systems
- Open Access
- 28 June 2021
- Published by Springer Nature in OR Spectrum
- Vol. 43 (3) , 603-605
Machine learning and combinatorial optimization, editorial
- 28 June 2021
- Published by Springer Nature in OR Spectrum
- Vol. 43 (3) , 603-605
Open AccessMachine learning and combinatorial optimization, editorial
- Open Access
- 30 April 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 67327-67348
Spatially-Distributed Missions With Heterogeneous Multi-Robot Teams
This work is about mission planning in teams of mobile autonomous agents. We consider tasks that are spatially distributed, non-atomic, and provide an utility for integral and also partial task completion. Agents are heterogeneous, therefore showing different efficiency when dealing with the tasks. The goal is to define a system-level plan that assigns tasks to agents to maximize mission performance. We define the mission planning problem through a model including multiple sub-problems that are addressed jointly: task selection and allocation, task scheduling, task routing, control of agent proximity over time. The problem is proven to be NP-hard and is formalized as a mixed integer linear program (MILP). Two solution approaches are proposed: one heuristic and one exact method. Both combine a generic MILP solver and a genetic algorithm, resulting in efficient anytime algorithms. To support performance scalability and to allow the effective use of the model when online continual replanning is required, a decentralized and fully distributed architecture is defined top-down from the MILP model. Decentralization drastically reduces computational requirements and shows good scalability at the expenses of only moderate losses in performance. Lastly, we illustrate the application of the mission planning framework in two demonstrators. These implementations show how the framework can be successfully integrated with different platforms, including mobile robots (ground and aerial), wearable computers, and smart-phone devices.
- 30 April 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 67327-67348
Open AccessSpatially-Distributed Missions With Heterogeneous Multi-Robot Teams
This work is about mission planning in teams of mobile autonomous agents. We consider tasks that are spatially distributed, non-atomic, and provide an utility for integral and also partial task completion. Agents are heterogeneous, therefore showing different efficiency when dealing with the tasks. The goal is to define a system-level plan that assigns tasks to agents to maximize mission performance. We define the mission planning problem through a model including multiple sub-problems that are addressed jointly: task selection and allocation, task scheduling, task routing, control of agent proximity over time. The problem is proven to be NP-hard and is formalized as a mixed integer linear program (MILP). Two solution approaches are proposed: one heuristic and one exact method. Both combine a generic MILP solver and a genetic algorithm, resulting in efficient anytime algorithms. To support performance scalability and to allow the effective use of the model when online continual replanning is required, a decentralized and fully distributed architecture is defined top-down from the MILP model. Decentralization drastically reduces computational requirements and shows good scalability at the expenses of only moderate losses in performance. Lastly, we illustrate the application of the mission planning framework in two demonstrators. These implementations show how the framework can be successfully integrated with different platforms, including mobile robots (ground and aerial), wearable computers, and smart-phone devices.
- Open Access
- 1 January 2021
- Published by INSTICC
- p. 514-522
Data Sharing and Assimilation in Multi-Robot Systems for Environment Mapping
- 1 January 2021
- Published by INSTICC
- p. 514-522
Open AccessData Sharing and Assimilation in Multi-Robot Systems for Environment Mapping
- 17 July 2019
- Published by Springer Nature
- p. 235-247
Robot Path Planning Using Imprecise and Sporadic Advisory Information from Humans
- 17 July 2019
- Published by Springer Nature
- p. 235-247
Robot Path Planning Using Imprecise and Sporadic Advisory Information from Humans
- Open Access
- 2 July 2018
- Published by Association for Computing Machinery (ACM)
- p. 21-28
A model of artificial emotions for behavior-modulation and implicit coordination in multi-robot systems
- 2 July 2018
- Published by Association for Computing Machinery (ACM)
- p. 21-28
Open AccessA model of artificial emotions for behavior-modulation and implicit coordination in multi-robot systems