Field of the Invention:
The present invention relates to the field of task scheduling and workload optimization in computational and organizational environments. Specifically, it integrates Ant Colony Optimization (ACO) and Reinforcement Learning (RL) techniques, including Q-Learning and SARSA, to develop an intelligent task scheduling and assignment system. The invention is applicable to cloud computing, distributed computing, industrial workflow automation, and large-scale resource management, providing adaptive, self-optimizing workload distribution strategies to improve task execution efficiency, resource utilization, and system performance
in dynamic environments..

Background of the proposed invention:
Efficient task scheduling and assignment are critical challenges in computational systems, cloud computing, distributed networks, and industrial workflow automation. As modem systems handle increasing workloads, traditional scheduling methods based on static rules or fixed priority algorithms struggle to balance task execution efficiency, resource utilization, and response time. These limitations lead to bottlenecks, inefficient resource allocation, and increased latency, particularly in dynamic and unpredictable environments where task demands
fluctuate.
Heuristic_algorithms,-such-as~Ant-Golony-Gptimization~("ACO)~have been widely used foF

scheduling due to their ability to find near-optimal solutions through pheromone-based pathfinding mechanisms. However, heuristic methods alone often lack adaptability to real-time changes in system conditions. Meanwhile, Reinforcement Learning (RL) approaches, including Q-Leaming and SARSA, provide a self-learning framework that allows scheduling systems to dynamically improve their deci si on-making processes by learning from past task execution experiences. Q-Leaming, an off-policy RL algorithm, optimizes long-term scheduling efficiency by evaluating the best possible action regardless of current policy constraints. SARSA, an on-policy RL algorithm, ensures stable and consistent learning in
environments with fluctuating workloads.

By integrating ACO for initial task allocation and RL-based optimization for continuous learning and adaptation, the proposed system addresses the limitations of traditional scheduling techniques by ensuring real-time, adaptive task distribution while maximizing system performance. This hybrid ACO-RL approach is particularly valuable for cloud computing platforms, industrial automation, and large-scale distributed systems, where intelligent workload optimization can significantly enhance task throughput, reduce latency, and improve
overall system efficiency.

Summary of the proposed-invention

The proposed invention introduces an intelligent task scheduling and assignment system that integrates Ant Colony Optimization (ACO) and Reinforcement Learning (RL) techniques, specifically Q-Leaming and SARSA, to optimize workload distribution in computational and
organizational environments.
The system leverages ACO’s pheromone-based heuristic optimization to generate efficient initial task schedules, while Q-Leaming and SARSA continuously refine scheduling strategies based on real-time system performance. The invention comprises key modules, including:
• Task Prioritization Module - Uses ACO to rank and allocate tasks based on urgency
and resource availability.

• RL-Based Task Assignment Engine - Employs Q-Leaming (off-policy) and SARSA (on-policy) to dynamically optimize task execution and resource distribution. • Workload Optimization Mechanism - Utilizes Markov Decision Processes (MDP) and adaptive learning to enhance task scheduling efficiency.
• Feedback Learning System - Continuously updates scheduling policies based on task completion time, system latency, and workload balancing metrics.
By combining heuristic search with adaptive learning, the system improves execution speed, enhances resource utilization, and reduces scheduling inefficiencies in dynamic environments.
This invention is particularly useful for cloud computing, distributed systems, industrial work flow automation, and large-scale resource management, ensuring real-time, intelligent
workload optimization for complex operational tasks.

Brief description of the proposed invention:
The proposed invention is an Al-driven task scheduling and assignment system that integrates Ant Colony Optimization (ACO) and Reinforcement Learning (RL) techniques, specifically Q-Learning and SARSA, to optimize workload distribution in computational environments.
The system utilizes ACO’s pheromone-based heuristic optimization to generate efficient initial task schedules, while Q-Leaming and SARSA continuously refine scheduling strategies based
on real-time system performance.

The system consists of four main components:
1. Task Prioritization Module - Uses ACO to evaluate and rank tasks based on urgency,
resource demand, and execution constraints.
2. RL-Based Task Assignment Engine - Implements Q-Learning for long-term policy optimization and SARSA for stable on-policy learning, ensuring dynamic and adaptive
task allocation.
3. Workload Optimization Mechanism - Employs Markov Decision Processes (MDP) to model task execution states and optimize decision-making for efficient wokload
balancing.

4. Feedback Learning System - Continuously updates task scheduling policies based on execution outcomes, resource utilization, and system performance metrics.
This hybrid ACO-RL scheduling framework ensures adaptive, self-optimizing workload distribution, reducing execution time, enhancing resource utilization, and improving overall system efficiency. The invention is particularly beneficial for cloud computing, distributed systems, industrial workflow automation, and large-scale operational management, providing a scalable, intelligent task scheduling solution for complex computing environments.

1. An Al-driven task scheduling and assignment system, comprising:
02-Apr-2025/33116/202541032559/Form 2(Title Page)
PATPN T
o A Task Prioritization Module utilizing Ant Colony Optimization (ACO) to rank tasks based on urgency, resource demand, and execution constraints;
o A Reinforcement Learning (RL)-Based Task Assignment Engine, integrating Q- Leaming and SARSA algorithms to optimize task allocation dynamically;
o A Workload Optimization Mechanism using Markov Decision Processes (MDP) to balance task execution across available resources;

o A Feedback Learning System that continuously refines scheduling policies based on execution outcomes, system latency, and workload distribution; and
o A Real-Time Monitoring and Adaptation Unit to analyze system performance
and adjust scheduling strategies accordingly.
2. The system of claim 1, wherein the Task Prioritization Module employs pheromone­based heuristic optimization from ACO to generate an initial task scheduling sequence.
3. The system of claim 1, wherein the RL-Based Task Assignment Engine uses Q-Learning’s off-policy learning approach to explore multiple task assignment
possibilities and optimize long-term efficiency.

4. The system of claim l, wherein the RL-Based Task Assignment Engine applies SARSA’s on-policy learning to ensure stable and adaptive decision-making in dynamic
workload environments.
5. The system of claim I, wherein the Workload Optimization Mechanism leverages Markov Decision Processes (MDP) to model task execution states and optimize
scheduling strategies based on reward functions.
6. The system of claim l , wherein the Feedback Learning System integrates reinforcement learning reward mechanisms to improve scheduling accuracy and efficiency over time.

77“ The system _of“claim 1, wherein the Real-Time Monitoring and Adaptation Unit analyzes task execution performance metrics and automatically adjusts task
assignments based on system load.
8. The system of claim 1, wherein the system dynamically updates its RL models based on newly acquired execution data, ensuring continuous improvement in scheduling
accuracy.
9. The system of claim 1, wherein the system includes a resource-aware load balancing module, ensuring efficient distribution of tasks across heterogeneous computing
environments.