Swarup Kumar Mohalik
Swarup Kumar Mohalik has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Ansuman Banerjee
Advanced Computing and Microelectronics Unit, Indian Statistical Institute, Kolkata, West Bengal, India
4 shared publications - Ramamurthy Badrinath
Ericsson Research Bangalore India
2 shared publications - Rupak Majumdar
Max Planck Institute for Software Systems (MPI-SWS), Kaiserslautern, Germany
2 shared publications A. Dury
Comput. Res. Inst. of Montreal (CRIM), Montreal, QC, Canada
1 shared publication- Ajay Kattepur
Ericsson Research, Bengaluru, India
1 shared publication - Alexandre Petrenko
CRIM
1 shared publication
Latest Publications
- Open Access
- 4 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 40-41
Towards Verification of Spiking Neural Networks for Next-Generation AI Systems
- 4 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 40-41
Open AccessTowards Verification of Spiking Neural Networks for Next-Generation AI Systems
- Open Access
- 13 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 40-50
Modeling and Verification of Sigma Delta Neural Networks using Satisfiability Modulo Theory
- 13 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 40-50
Open AccessModeling and Verification of Sigma Delta Neural Networks using Satisfiability Modulo Theory
- Open Access
- 6 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Integrated Task and Path Planning for Collaborative Multi-Robot Systems
- 6 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Open AccessIntegrated Task and Path Planning for Collaborative Multi-Robot Systems
- Open Access
Configuring Safe Spiking Neural Controllers for Cyber-Physical Systems through Formal Verification
Spiking Neural Networks (SNNs) are a subclass of neuromorphic models that have great potential to be used as controllers in Cyber-Physical Systems (CPSs) due to their energy efficiency. They can benefit from the prevalent approach of first training an Artificial Neural Network (ANN) and then translating to an SNN with subsequent hyperparameter tuning. The tuning is required to ensure that the resulting SNN is accurate with respect to the ANN in terms of metrics like Mean Squared Error (MSE). However, SNN controllers for safety-critical CPSs must also satisfy safety specifications, which are not guaranteed by the conversion approach. In this paper, we propose a solution which tunes the $temporal$ $window$ hyperparameter of the translated SNN to ensure both accuracy and compliance with the safe range specification that requires the SNN outputs to remain within a safe range. The core verification problem is modelled using mixed-integer linear programming (MILP) and is solved with Gurobi. When the controller fails to meet the range specification, we compute tight bounds on the SNN outputs as feedback for the CPS developer. To mitigate the high computational cost of verification, we integrate data-driven steps to minimize verification calls. Our approach provides designers with the confidence to safely integrate energy-efficient SNN controllers into modern CPSs. We demonstrate our approach with experimental results on five different benchmark neural controllers.Open AccessConfiguring Safe Spiking Neural Controllers for Cyber-Physical Systems through Formal Verification
Spiking Neural Networks (SNNs) are a subclass of neuromorphic models that have great potential to be used as controllers in Cyber-Physical Systems (CPSs) due to their energy efficiency. They can benefit from the prevalent approach of first training an Artificial Neural Network (ANN) and then translating to an SNN with subsequent hyperparameter tuning. The tuning is required to ensure that the resulting SNN is accurate with respect to the ANN in terms of metrics like Mean Squared Error (MSE). However, SNN controllers for safety-critical CPSs must also satisfy safety specifications, which are not guaranteed by the conversion approach. In this paper, we propose a solution which tunes the $temporal$ $window$ hyperparameter of the translated SNN to ensure both accuracy and compliance with the safe range specification that requires the SNN outputs to remain within a safe range. The core verification problem is modelled using mixed-integer linear programming (MILP) and is solved with Gurobi. When the controller fails to meet the range specification, we compute tight bounds on the SNN outputs as feedback for the CPS developer. To mitigate the high computational cost of verification, we integrate data-driven steps to minimize verification calls. Our approach provides designers with the confidence to safely integrate energy-efficient SNN controllers into modern CPSs. We demonstrate our approach with experimental results on five different benchmark neural controllers. - Open Access
Optimal Integrated Task and Path Planning and Its Application to Multi-Robot Pickup and Delivery
We propose a generic multi-robot planning mechanism that combines an optimal task planner and an optimal path planner to provide a scalable solution for complex multi-robot planning problems. The Integrated planner, through the interaction of the task planner and the path planner, produces optimal collision-free trajectories for the robots. We illustrate our general algorithm on an object pick-and-drop planning problem in a warehouse scenario where a group of robots is entrusted with moving objects from one location to another in the workspace. We solve the task planning problem by reducing it into an SMT-solving problem and employing the highly advanced SMT solver Z3 to solve it. To generate collision-free movement of the robots, we extend the state-of-the-art algorithm Conflict Based Search with Precedence Constraints with several domain-specific constraints. We evaluate our integrated task and path planner extensively on various instances of the object pick-and-drop planning problem and compare its performance with a state-of-the-art multi-robot classical planner. Experimental results demonstrate that our planning mechanism can deal with complex planning problems and outperforms a state-of-the-art classical planner both in terms of computation time and the quality of the generated plan.
Open AccessOptimal Integrated Task and Path Planning and Its Application to Multi-Robot Pickup and Delivery
We propose a generic multi-robot planning mechanism that combines an optimal task planner and an optimal path planner to provide a scalable solution for complex multi-robot planning problems. The Integrated planner, through the interaction of the task planner and the path planner, produces optimal collision-free trajectories for the robots. We illustrate our general algorithm on an object pick-and-drop planning problem in a warehouse scenario where a group of robots is entrusted with moving objects from one location to another in the workspace. We solve the task planning problem by reducing it into an SMT-solving problem and employing the highly advanced SMT solver Z3 to solve it. To generate collision-free movement of the robots, we extend the state-of-the-art algorithm Conflict Based Search with Precedence Constraints with several domain-specific constraints. We evaluate our integrated task and path planner extensively on various instances of the object pick-and-drop planning problem and compare its performance with a state-of-the-art multi-robot classical planner. Experimental results demonstrate that our planning mechanism can deal with complex planning problems and outperforms a state-of-the-art classical planner both in terms of computation time and the quality of the generated plan.
- Open Access
- 25 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-8
CONRAD: Cognitive Intent Driven 5G Network Slice Planning and Design
The emergence of 5G allows the network to be sliced to meet multiple service requirements. However, current design systems are policy-driven, which limits the ability to redesign the 5G network adaptively for requirements or traffic patterns. In this paper, we propose Conrad, an autonomous and intent-driven framework that can allocate resources efficiently to meet 5G slice requirements. Via the use of a cognitive intent handling engine, the requirements are transformed to goals that may be solved using machine reasoning techniques. Four agents are registered for forecasting, design planning, evaluation and actuation to enable dynamic re-design of the 5G slices. Exploiting automated planning techniques, proposals are provided to optimally provision the network slice resources. This process is demonstrated over a radio access network (RAN) slicing use case with intents on average expected throughput.
- 25 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessCONRAD: Cognitive Intent Driven 5G Network Slice Planning and Design
The emergence of 5G allows the network to be sliced to meet multiple service requirements. However, current design systems are policy-driven, which limits the ability to redesign the 5G network adaptively for requirements or traffic patterns. In this paper, we propose Conrad, an autonomous and intent-driven framework that can allocate resources efficiently to meet 5G slice requirements. Via the use of a cognitive intent handling engine, the requirements are transformed to goals that may be solved using machine reasoning techniques. Four agents are registered for forecasting, design planning, evaluation and actuation to enable dynamic re-design of the 5G slices. Exploiting automated planning techniques, proposals are provided to optimally provision the network slice resources. This process is demonstrated over a radio access network (RAN) slicing use case with intents on average expected throughput.
- Open Access
HAMMER: Multi-Level Coordination of Reinforcement Learning Agents via Learned Messaging
Cooperative multi-agent reinforcement learning (MARL) has achieved significant results, most notably by leveraging the representation-learning abilities of deep neural networks. However, large centralized approaches quickly become infeasible as the number of agents scale, and fully decentralized approaches can miss important opportunities for information sharing and coordination. Furthermore, not all agents are equal -- in some cases, individual agents may not even have the ability to send communication to other agents or explicitly model other agents. This paper considers the case where there is a single, powerful, \emph{central agent} that can observe the entire observation space, and there are multiple, low-powered \emph{local agents} that can only receive local observations and are not able to communicate with each other. The central agent's job is to learn what message needs to be sent to different local agents based on the global observations, not by centrally solving the entire problem and sending action commands, but by determining what additional information an individual agent should receive so that it can make a better decision. In this work we present our MARL algorithm \algo, describe where it would be most applicable, and implement it in the cooperative navigation and multi-agent walker domains. Empirical results show that 1) learned communication does indeed improve system performance, 2) results generalize to heterogeneous local agents, and 3) results generalize to different reward structures.
Open AccessHAMMER: Multi-Level Coordination of Reinforcement Learning Agents via Learned Messaging
Cooperative multi-agent reinforcement learning (MARL) has achieved significant results, most notably by leveraging the representation-learning abilities of deep neural networks. However, large centralized approaches quickly become infeasible as the number of agents scale, and fully decentralized approaches can miss important opportunities for information sharing and coordination. Furthermore, not all agents are equal -- in some cases, individual agents may not even have the ability to send communication to other agents or explicitly model other agents. This paper considers the case where there is a single, powerful, \emph{central agent} that can observe the entire observation space, and there are multiple, low-powered \emph{local agents} that can only receive local observations and are not able to communicate with each other. The central agent's job is to learn what message needs to be sent to different local agents based on the global observations, not by centrally solving the entire problem and sending action commands, but by determining what additional information an individual agent should receive so that it can make a better decision. In this work we present our MARL algorithm \algo, describe where it would be most applicable, and implement it in the cooperative navigation and multi-agent walker domains. Empirical results show that 1) learned communication does indeed improve system performance, 2) results generalize to heterogeneous local agents, and 3) results generalize to different reward structures.
- 8 June 2021
- Published by River Publishers in Journal of ICT Standardization
- p. 257–290
Intent-driven Closed Loops for Autonomous Networks
Closed loops are key enablers for automation that have been successfully used in many industries for long, and more recently for computing and networking applications. The Zero-touch network and service management (ZSM) framework introduced standardized components that allow the creation, execution, and governance of multiple closed loops, enabling zero-touch management of end-to-end services across different management domains. However, the coordinated and optimal instantiation and operation of multiple closed loops is an open question that is left for implementation by the ZSM specifications. In this paper, we propose a methodology that uses intents as a way of communicating requirements to be considered by autonomous management domains to coordinate hierarchies of closed loops. The intent-driven methodology facilitates hierarchical and peer interactions for delegation and escalation of intents. Furthermore, it extends the existing management capabilities of the ZSM framework and facilitates conflict-free integration of closed loops by setting optimal (and non-conflicting) goals that each closed loop in the hierarchy needs to account for. We show an example of the application of the proposed methodology in a network slicing assurance use case. The new capabilities introduced in this paper can be considered as an extension of the ZSM framework to be used in scenarios where multiple intent-driven closed loops exist.
- 8 June 2021
- Published by River Publishers in Journal of ICT Standardization
- p. 257–290
Intent-driven Closed Loops for Autonomous Networks
Closed loops are key enablers for automation that have been successfully used in many industries for long, and more recently for computing and networking applications. The Zero-touch network and service management (ZSM) framework introduced standardized components that allow the creation, execution, and governance of multiple closed loops, enabling zero-touch management of end-to-end services across different management domains. However, the coordinated and optimal instantiation and operation of multiple closed loops is an open question that is left for implementation by the ZSM specifications. In this paper, we propose a methodology that uses intents as a way of communicating requirements to be considered by autonomous management domains to coordinate hierarchies of closed loops. The intent-driven methodology facilitates hierarchical and peer interactions for delegation and escalation of intents. Furthermore, it extends the existing management capabilities of the ZSM framework and facilitates conflict-free integration of closed loops by setting optimal (and non-conflicting) goals that each closed loop in the hierarchy needs to account for. We show an example of the application of the proposed methodology in a network slicing assurance use case. The new capabilities introduced in this paper can be considered as an extension of the ZSM framework to be used in scenarios where multiple intent-driven closed loops exist.
- 22 April 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Services Computing
- Vol. 15 (5) , 3031-3046
Service Selection With Package Bundles and Compatibility Constraints
With the rapid proliferation of strategic alliances between service providers, enterprises cooperate towards service quality improvement abd provide lower cost service bundles. This paper presents a novel solution to the minimum cost service bundle selection problem for workflows in the presence of singleton subscription costs and service bundle offerings and compatibility requirements. Given a workflow specifying a set of tasks and a set of candidate services for each task, with a set of compatibility constraints between services, the selection problem has the objective of selecting the most suitable service offering(s) for each task. In this paper, we analyze the selection problem in the presence of service bundle offerings. We present a novel multi-partite hyper-graph visualization of the selection problem and analyze its hardness. Additionally we present a novel combination of ILP and abstraction refinement as a potential solution, that is shown to expedite a na\"ive ILP based solution. We present experiments to substantiate this claim.
- 22 April 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Services Computing
- Vol. 15 (5) , 3031-3046
Service Selection With Package Bundles and Compatibility Constraints
With the rapid proliferation of strategic alliances between service providers, enterprises cooperate towards service quality improvement abd provide lower cost service bundles. This paper presents a novel solution to the minimum cost service bundle selection problem for workflows in the presence of singleton subscription costs and service bundle offerings and compatibility requirements. Given a workflow specifying a set of tasks and a set of candidate services for each task, with a set of compatibility constraints between services, the selection problem has the objective of selecting the most suitable service offering(s) for each task. In this paper, we analyze the selection problem in the presence of service bundle offerings. We present a novel multi-partite hyper-graph visualization of the selection problem and analyze its hardness. Additionally we present a novel combination of ILP and abstraction refinement as a potential solution, that is shown to expedite a na\"ive ILP based solution. We present experiments to substantiate this claim.
- Open Access
Smart Contracts for Multiagent Plan Execution in Untrusted Cyber-physical Systems
Intelligent Cyber-physical systems can be modelled as multi-agent systems with planning capability to impart adaptivity for changing contexts. In such multi-agent systems, the protocol for plan execution must result in the proper completion and ordering of actions in spite of their distributed execution. However, in untrusted scenarios, there is a possibility of agents not respecting the protocol either due to faults or due to malicious reasons thereby resulting in plan failure. In order to prevent such situations, we propose to implement the execution of agents through smart contracts. This points to a generic architecture seamlessly integrating intelligent planning-based CPS and smart-contracts.
Open AccessSmart Contracts for Multiagent Plan Execution in Untrusted Cyber-physical Systems
Intelligent Cyber-physical systems can be modelled as multi-agent systems with planning capability to impart adaptivity for changing contexts. In such multi-agent systems, the protocol for plan execution must result in the proper completion and ordering of actions in spite of their distributed execution. However, in untrusted scenarios, there is a possibility of agents not respecting the protocol either due to faults or due to malicious reasons thereby resulting in plan failure. In order to prevent such situations, we propose to implement the execution of agents through smart contracts. This points to a generic architecture seamlessly integrating intelligent planning-based CPS and smart-contracts.