Sukrit Mittal
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Interests
Sukrit Mittal has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Dhish Kumar Saxena
Indian Institute of Technology Roorkee
6 shared publications - Kalyanmoy Deb
Department of Computer and Electrical Engineering, Michigan State University, USA and Department of Computer Science, University of Alabama, Tuscaloosa, USA
5 shared publications - Erik David Goodman
Michigan State University
3 shared publications - Daniel Ostrov
Santa Clara University
2 shared publications - Harshad Khadilkar
Indian Institute of Technology, Bombay, Mumbai, Maharashtra, India
2 shared publications Rishabh Singh
Mahindra Mahindra, Ltd
2 shared publications
Latest Publications
- Open Access
- 1 December 2026
- Published by KEAI Publishing in The Journal of Finance and Data Science
- Vol. 12
A meta reinforcement learning approach to goals-based wealth management
- 1 December 2026
- Published by KEAI Publishing in The Journal of Finance and Data Science
- Vol. 12
Open AccessA meta reinforcement learning approach to goals-based wealth management
- Open Access
- 30 October 2026
- Published by Elsevier in SSRN Electronic Journal
CARL: Chat Assisted Reinforcement Learning
Failure of applied Reinforcement Learning (RL) to solve a given problem predominantly stems from unaddressed "specification gaps," where the intent of human experts in the problem's domain is inaccurately or incompletely translated by AI into mathematical objectives, constraints, and reward functions. Failure can also stem from unaddressed "performance gaps," where human domain experts are unable to determine the validity of the RL based solution or where parameters (i.e., hyperparameters or weights on the reward terms comprising the reward) are poorly chosen. Chat Assisted Reinforcement Learning (CARL) bridges these gaps through a human-AI co-design interface architecture that enforces interactive specification prior to code generation and then enables subsequent verification and parameter alterations via iteration. More specifically, CARL leads domain experts through a structured ten-step workflow for each iteration using step-specific skill files, domain-aware suggestions via an assistant persona (called Trillian), and compact decision summaries that maintain context across iterations without context overload. Once the domain expert confirms the Markov decision process (MDP) specified by this procedure, CARL's autonomous tool-use loop generates, tests, executes, and debugs an end-to-end RL training pipeline, followed by an interactive post-training diagnosis and revisit step that allows for verification and parameter alteration before running a new iteration if desired. We apply CARL to two disparate, path-dependent optimization domains: (1) goals-based wealth management in finance and (2) dynamic-pressure-constrained Goddard rocket control in aerospace engineering. In both of these examples, CARL demonstrates that interactive specification engineering using human domain expertise throughout the process resolves critical failure modes (such as reward hacking and state aliasing) and consistently produces viable policies where unguided, fully autonomous code generation fails. Full implementation available at https://github.com/srdas/carl.
- 30 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCARL: Chat Assisted Reinforcement Learning
Failure of applied Reinforcement Learning (RL) to solve a given problem predominantly stems from unaddressed "specification gaps," where the intent of human experts in the problem's domain is inaccurately or incompletely translated by AI into mathematical objectives, constraints, and reward functions. Failure can also stem from unaddressed "performance gaps," where human domain experts are unable to determine the validity of the RL based solution or where parameters (i.e., hyperparameters or weights on the reward terms comprising the reward) are poorly chosen. Chat Assisted Reinforcement Learning (CARL) bridges these gaps through a human-AI co-design interface architecture that enforces interactive specification prior to code generation and then enables subsequent verification and parameter alterations via iteration. More specifically, CARL leads domain experts through a structured ten-step workflow for each iteration using step-specific skill files, domain-aware suggestions via an assistant persona (called Trillian), and compact decision summaries that maintain context across iterations without context overload. Once the domain expert confirms the Markov decision process (MDP) specified by this procedure, CARL's autonomous tool-use loop generates, tests, executes, and debugs an end-to-end RL training pipeline, followed by an interactive post-training diagnosis and revisit step that allows for verification and parameter alteration before running a new iteration if desired. We apply CARL to two disparate, path-dependent optimization domains: (1) goals-based wealth management in finance and (2) dynamic-pressure-constrained Goddard rocket control in aerospace engineering. In both of these examples, CARL demonstrates that interactive specification engineering using human domain expertise throughout the process resolves critical failure modes (such as reward hacking and state aliasing) and consistently produces viable policies where unguided, fully autonomous code generation fails. Full implementation available at https://github.com/srdas/carl.
- Open Access
- 14 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1025-1041
Machine Learning Assisted Evolutionary Multi-objective Optimization
- 14 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1025-1041
Open AccessMachine Learning Assisted Evolutionary Multi-objective Optimization
- 2 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 28 (6) , 1605-1619
A Unified Innovized Progress Operator for Performance Enhancement in Evolutionary Multi- and Many-Objective Optimization
This paper proposes a machine learning (ML) based unified innovized progress (UIP) operator to simultaneously enhance the convergence and diversity capabilities of reference vector based evolutionary multi-and many-objective optimization algorithms, namely, RV-EMâOAs. Recent studies have demonstrated that ML intervention could help enhance convergence of RV-EMâOAs by capturing efficient search directions, through mapping of inter-generational solutions along the different reference vectors (RVs). This paper first demonstrates that ML intervention can also help enhance the diversity capability of RV-EMâOAs through mapping of intra-generational solutions across the RVs. Subsequently, the UIP operator integrates the convergence and diversity enhancement capabilities in a manner that is generic -applicable to different RV-EMâOAs, and practicable -not requiring any extra solution evaluations over the base RV-EMâOAs. Based on 24,056 experimental runs on multi-and many-objective problems, the UIP operator, when integrated with different RV-EMâOAs, has provided statistically better performance in about 36% instances, and better or equivalent in about 92% instances, compared to the respective base RV-EMâOAs.
- 2 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 28 (6) , 1605-1619
A Unified Innovized Progress Operator for Performance Enhancement in Evolutionary Multi- and Many-Objective Optimization
This paper proposes a machine learning (ML) based unified innovized progress (UIP) operator to simultaneously enhance the convergence and diversity capabilities of reference vector based evolutionary multi-and many-objective optimization algorithms, namely, RV-EMâOAs. Recent studies have demonstrated that ML intervention could help enhance convergence of RV-EMâOAs by capturing efficient search directions, through mapping of inter-generational solutions along the different reference vectors (RVs). This paper first demonstrates that ML intervention can also help enhance the diversity capability of RV-EMâOAs through mapping of intra-generational solutions across the RVs. Subsequently, the UIP operator integrates the convergence and diversity enhancement capabilities in a manner that is generic -applicable to different RV-EMâOAs, and practicable -not requiring any extra solution evaluations over the base RV-EMâOAs. Based on 24,056 experimental runs on multi-and many-objective problems, the UIP operator, when integrated with different RV-EMâOAs, has provided statistically better performance in about 36% instances, and better or equivalent in about 92% instances, compared to the respective base RV-EMâOAs.
- 4 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 27 (4) , 923-937
A Localized High-Fidelity-Dominance-Based Many-Objective Evolutionary Algorithm
The practicality of Pareto-dominance in solving many-objective optimization problems becomes questionable due to its inability to factor the critical Human Decision-making (HDM) elements, including, the number of better objectives, the degree of betterment in objectives, and objectives’ relative preference. Relevant dominance principles are recently proposed to incorporate the first two HDM elements, often with the need for new tunable parameters. This paper proposes a high-fidelity-dominance principle, that factors all the three HDM elements, explicitly and simultaneously, and without requiring tuning of any parameter. This principle has been implemented in a reference vector based framework, leading to a computationally efficient many-objective Evolutionary Algorithm (MaOEA), namely localized high-fidelity-dominance based EA (LHFiD). Critically, LHFiD also has an inbuilt mechanism for on-the-fly determination of the timing for: (a) intermittent nadir point estimation that enables faster convergence, and (b) its self-termination that bears practically utility. This paper is based on an extensive study involving 41,912 experiments, in which the proposed LHFiD approach is compared with existing competitive MaOEAs. The paper reports statistically better performance in about 60% instances, making it practical and worthy of further investigation and application.
- 4 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 27 (4) , 923-937
A Localized High-Fidelity-Dominance-Based Many-Objective Evolutionary Algorithm
The practicality of Pareto-dominance in solving many-objective optimization problems becomes questionable due to its inability to factor the critical Human Decision-making (HDM) elements, including, the number of better objectives, the degree of betterment in objectives, and objectives’ relative preference. Relevant dominance principles are recently proposed to incorporate the first two HDM elements, often with the need for new tunable parameters. This paper proposes a high-fidelity-dominance principle, that factors all the three HDM elements, explicitly and simultaneously, and without requiring tuning of any parameter. This principle has been implemented in a reference vector based framework, leading to a computationally efficient many-objective Evolutionary Algorithm (MaOEA), namely localized high-fidelity-dominance based EA (LHFiD). Critically, LHFiD also has an inbuilt mechanism for on-the-fly determination of the timing for: (a) intermittent nadir point estimation that enables faster convergence, and (b) its self-termination that bears practically utility. This paper is based on an extensive study involving 41,912 experiments, in which the proposed LHFiD approach is compared with existing competitive MaOEAs. The paper reports statistically better performance in about 60% instances, making it practical and worthy of further investigation and application.
- 1 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 26 (5) , 961-975
Enhanced Innovized Progress Operator for Evolutionary Multi- and Many-Objective Optimization
Innovization is a task of learning common relationships among some or all of the Pareto-optimal (PO) solutions in multi-and many-objective optimization problems. A recent study has shown that a chronological sequence of non-dominated solutions obtained along the successive generations of an optimizer possesses salient patterns that can be learnt using a Machine Learning (ML) model, and can help the offspring solutions progress in useful directions. This paper enhances each constitutive module of the above approach, including novel interventions on management of the convergence-diversity tradeoff while mapping the solutions from the previous and current generation; use of a computationally more efficient ML method, namely Random Forest; and changing the manner and extent to which the learnt ML model is utilized towards advancement of the offspring. The proposed modules constitute what is called the enhanced innovized progress (IP2) operator. To investigate the search efficacy provided by the IP2 operator, it is integrated with multi-and many-objective optimization algorithms, such as NSGA-II, NSGA-III, MOEA/D, and MaOEA-IGD, and tested on a range of two-to ten-objective test problems, and five real-world problems. Since the IP2 operator utilizes the history of gradual and progressive improvements in solutions over generations, without requiring any additional solution evaluations, it opens up a new direction for ML-assisted evolutionary optimization.
- 1 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 26 (5) , 961-975
Enhanced Innovized Progress Operator for Evolutionary Multi- and Many-Objective Optimization
Innovization is a task of learning common relationships among some or all of the Pareto-optimal (PO) solutions in multi-and many-objective optimization problems. A recent study has shown that a chronological sequence of non-dominated solutions obtained along the successive generations of an optimizer possesses salient patterns that can be learnt using a Machine Learning (ML) model, and can help the offspring solutions progress in useful directions. This paper enhances each constitutive module of the above approach, including novel interventions on management of the convergence-diversity tradeoff while mapping the solutions from the previous and current generation; use of a computationally more efficient ML method, namely Random Forest; and changing the manner and extent to which the learnt ML model is utilized towards advancement of the offspring. The proposed modules constitute what is called the enhanced innovized progress (IP2) operator. To investigate the search efficacy provided by the IP2 operator, it is integrated with multi-and many-objective optimization algorithms, such as NSGA-II, NSGA-III, MOEA/D, and MaOEA-IGD, and tested on a range of two-to ten-objective test problems, and five real-world problems. Since the IP2 operator utilizes the history of gradual and progressive improvements in solutions over generations, without requiring any additional solution evaluations, it opens up a new direction for ML-assisted evolutionary optimization.
- 24 March 2021
- Published by Springer Nature
- p. 89-101
Embedding a Repair Operator in Evolutionary Single and Multi-objective Algorithms - An Exploitation-Exploration Perspective
- 24 March 2021
- Published by Springer Nature
- p. 89-101
Embedding a Repair Operator in Evolutionary Single and Multi-objective Algorithms - An Exploitation-Exploration Perspective
- 11 August 2020
- Published by Springer Nature
- p. 225-233
Innovative Design of Hydraulic Actuation System for Operator Fatigue Reduction and Its Optimization
- 11 August 2020
- Published by Springer Nature
- p. 225-233
Innovative Design of Hydraulic Actuation System for Operator Fatigue Reduction and Its Optimization
- 9 January 2019
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
A Simplified Approach towards Draft Control in Hydraulic Machines for Component/Cost Reduction
Draft Control is a mechanism in tractors to maintain the penetration of its rear attachment (plough) in ground to avoid engine overload and keep the operation under control. The concept was introduced in excavators for controlling the cutting edge penetration of the bucket in the ground. The idea is to implement draft control automation in excavators without involving complex or expensive electronic sensors and ECUs. The proposition is to use an adjustable pressure relief valve to sense the pressure acting in the cylinder. The relief valve acts as pilot control valve to control the Boom input, thus maintaining the digging depth of the excavator bucket. The relief valve adjustment allows the user to control the bucket through different terrains, i.e., hard debris, soft soil, dust, etc. thus allowing the engine and pump to run under control conditions at all times. The potential expands to the optimizing operational efficiency with minimalistic addition of components. The paper includes detailed concept and MATLAB model to validate the system.
- 9 January 2019
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
A Simplified Approach towards Draft Control in Hydraulic Machines for Component/Cost Reduction
Draft Control is a mechanism in tractors to maintain the penetration of its rear attachment (plough) in ground to avoid engine overload and keep the operation under control. The concept was introduced in excavators for controlling the cutting edge penetration of the bucket in the ground. The idea is to implement draft control automation in excavators without involving complex or expensive electronic sensors and ECUs. The proposition is to use an adjustable pressure relief valve to sense the pressure acting in the cylinder. The relief valve acts as pilot control valve to control the Boom input, thus maintaining the digging depth of the excavator bucket. The relief valve adjustment allows the user to control the bucket through different terrains, i.e., hard debris, soft soil, dust, etc. thus allowing the engine and pump to run under control conditions at all times. The potential expands to the optimizing operational efficiency with minimalistic addition of components. The paper includes detailed concept and MATLAB model to validate the system.
- 9 January 2019
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
A Simplified Method to Form a Fuel Economy Test Cycle for Test Tracks/Autonomous Vehicles
Right now, most of the industries are doing real world fuel economy test of their proto-types/mules by deploying them to city, state highways and national highways to declare their FE value. Which involves lot of unexpected problems, large amount of labor and the worst part is the pollution, traffic and uncontrolled randomness (difficulty in completing the test during festivals).Among already existing drive cycles, this is different because of a simplified mathematics of error and probabilities for method of cycle creation and it has been correlated with the real-world results.This paper reveals a 10-min city drive cycle which was created after studying most of the parameters of vehicle which may affect fuel economy and a whole method, which was devised to form this cycle and then to simplify it enough to drive.Using the derived method some standard cycles for different segments can be created by using already existing data (all industries have plethora of recorded data) and vehicle/engine specifications will be the deciding factor for which cycle needs to be followed.
- 9 January 2019
- Published by SAE International in SAE International Journal of Advances and Current Practices in Mobility
- Vol. 1
A Simplified Method to Form a Fuel Economy Test Cycle for Test Tracks/Autonomous Vehicles
Right now, most of the industries are doing real world fuel economy test of their proto-types/mules by deploying them to city, state highways and national highways to declare their FE value. Which involves lot of unexpected problems, large amount of labor and the worst part is the pollution, traffic and uncontrolled randomness (difficulty in completing the test during festivals).Among already existing drive cycles, this is different because of a simplified mathematics of error and probabilities for method of cycle creation and it has been correlated with the real-world results.This paper reveals a 10-min city drive cycle which was created after studying most of the parameters of vehicle which may affect fuel economy and a whole method, which was devised to form this cycle and then to simplify it enough to drive.Using the derived method some standard cycles for different segments can be created by using already existing data (all industries have plethora of recorded data) and vehicle/engine specifications will be the deciding factor for which cycle needs to be followed.