Welcome to my blog! I'm Ayesha Zulfiqar, and today I delve into two recent AI research papers that have deepened my understanding of intelligent agents and search algorithms—topics we explored in our AI course. I'll share insights from these papers and discuss what I found most intriguing.
Paper 1 : The Rise of Agentic AI
The very first paper I read is “The Rise of Agentic AI : A Review of Definitions, Frameworks, and Challenges.” This paper is about agentic AI, which means AI systems that can perform on their own and make decisions on their own rather than waiting for humans to give directions. They not only respond to input but also plan actions, interact with the environment, and complete tasks autonomously. A familiar example of agentic AI in the real world is self-driving cars. These vehicles are designed to navigate traffic, detect obstacles, and make split-second decisions on their own with minimal human input, showing how agentic AI is already shaping technologies we use or hear about every day.
This is exactly related to the concept of intelligent agents that we studied in our AI course. An agent perceives the environment using sensors and then performs actions on the environment through the actuators to achieve the goals. This paper says that modern AI is moving towards more complex, agent-based systems, but it also mentions challenges like safety, reliability, and control when deploying autonomous agents in real-world scenarios.
I found it very interesting that simple decisions in Agentic AI need complex rules to make sure nothing goes wrong. This shows that building autonomous systems is not only about making them smart, but also about making them safe and more reliable in different situations. For example, one big challenge the paper highlights is ensuring the safety of these systems. Even a small mistake by an autonomous agent, like a self-driving car misjudging an obstacle, could lead to serious real-world harm. This makes it crucial to carefully design and test these systems to prevent accidents and other unintended consequences. I also learned how many components are involved in an agentic system, such as planning, reasoning, and interaction with the environment.
- Research on the A* Algorithm with Adaptive Weights (2025)
The second research paper that i read is “Research on the A* Algorithm Based on Adaptive Weights and Heuristic Reward Values”.
This paper discusses improving the A* search algorithms which is commonly used in robotics, games and navigation systems to find shortest paths. This paper introduces several key improvements : Diagonal Free Five Way Search : Instead of checking 8 directions like normal A*, the algorithm only checks 5 directions and avoids diagonal moves when an obstacle is close. This helps to prevent the agent from clipping corners. Adaptive Dynamic Weighting : This algorithm uses a radial function to change the heuristic weight dynamically. Near the start, the weight is high to move faster towards the goal, then decreases to 1 as the agent gets closer, ensuring path accuracy. Heuristic Reward Values : The reward values are diffused from the goal backward to prevent the agent from getting stuck behind obstacles and guide it around complex barriers. Path Smoothing : Once the path is found, the extra nodes are removed, and the Bézier curves smooth the path. This increases the distance from the obstacles and makes the trajectory safer. The paper connects directly to what we learned in class about informed searches and heuristic value. While we studied the A* algorithm theoretically the research shows how adaptive weights and heuristic rewards can improve its efficiency in real world applications like robotics and navigation. Personal Insight The thing I found the most interesting was how the concepts of agents and search algorithms we learn in class are actually applied and enhanced in real research. Reading these papers helped me understand the flow of technical things and refine my understanding. I also took help from NotebookLLM, which made it easy to understand the complex sections.
I also explained both papers in a small video; you can check it here :
https://youtu.be/ujAFZD3Xg9I?si=yqwGm4_ko8gM7ngJ
These research papers showed me how AI concepts from our course are applied to create systems that are more efficient and reliable in the real world
Top comments (0)