Cognitive Computation and Systems

Fuchun Sun, Jianmin LiCommunications 2024 · DOI · OpenAlex

This document represents the collected proceedings of the Second International Conference on Cognitive Computation and Systems (ICCCS 2023). As part of the Communications in Computer and Information Science series, it provides a snapshot of the current state of "cognitive computing"—an interdisciplinary field dedicated to building intelligent systems that mimic human-like learning, reasoning, and interaction.

TL;DR

ICCCS 2023 presents 26 high-quality papers exploring how human-inspired cognitive models can be applied to complex engineering challenges. The volume spans two primary tracks: "Perception and Learning" and "Decision Making and Systems," covering everything from UAV swarm awareness to AI-driven medical diagnostics.

What Problem Does This Paper Address?

The core challenge addressed by this collection is the gap between traditional, rigid algorithmic processing and the flexible, nuanced nature of human cognition. While standard AI excels at specific tasks, "cognitive computing" aims to develop systems that can process and analyze massive amounts of complex, often "noisy" data in a way that mimics human thought processes.

The goal is to enable agents to understand the world more naturally, allowing them to solve problems that are currently too difficult for humans to handle alone or for traditional machines to navigate—such as predicting natural disasters, improving autonomous vehicle safety in unpredictable traffic, and enhancing human-robot collaboration.

Methods

Because this is a conference proceedings volume, the methods are diverse and represent the "cutting edge" of the field as of late 2023. Key technical approaches highlighted across the 26 papers include:

  • Prompt Learning: Using large models to handle task assignment for heterogeneous robots.
  • Meta-Reinforcement Learning: Applied to synthesize highway merging policies for automated vehicles with "latent context imagination."
  • Advanced Neural Architectures: Including Siamese Networks with attention mechanisms for drug combination prediction and Multi-Stage Temporal Convolutional Networks (MS-MH-TCN) for action segmentation.
  • Bio-Inspired Optimization: Such as Multi-population Fruit Fly Optimization for path planning.
  • Hybrid Systems: Combining fuzzy logic with top-down attention frameworks for selective observation.
  • Infrastructure & Security: Utilizing Blockchain and Edge Computing for intermodal transport and Byzantine Fault Tolerant algorithms for distributed systems.

Datasets

While the front matter does not list specific shared benchmarks, the individual papers utilize a wide array of domain-specific data types for training and evaluation, including:

  • Medical/Physiological: Tongue images for Traditional Chinese Medicine (TCM) classification and physiological signals for exoskeleton robot pain recognition.
  • Navigation/Vision: LiDAR/Laser ranging data for UAV target location and landmark recognition for visual-inertial navigation.
  • Industrial/Logistics: Data for intermodal transport sharing and license plate recognition.
  • Games: A specialized platform for "Tibetan Jiu Chess" intelligent gaming.

Key Insights and Ideas

  • Cognition as an Industrial Tool: The conference emphasizes that cognitive computing is moving out of pure theory and into industrial applications, particularly in autonomous vehicles, intelligent robots, and UAV swarms.
  • The Importance of "Attention": Several papers focus on "attention modulation" and "attention mechanisms," suggesting that mimicking the human ability to filter irrelevant information is a primary hurdle for modern perception systems.
  • Human-Machine Synergy: A recurring theme is the design of systems that interact with humans in a "more natural way," particularly in the context of exoskeleton robots and visually impaired assistance.
  • Decentralized Intelligence: The inclusion of blockchain and Byzantine fault tolerance papers suggests a growing interest in making cognitive systems not just "smart," but also secure and resilient in distributed environments.

Key Results

The conference maintained a competitive 38% acceptance rate (26 papers selected from 68 submissions), ensuring a high standard of peer-reviewed content.

  • Perception: Advances were reported in real-time anomaly detection for densely placed objects and robust license plate recognition using pre-training segmentation models.
  • Decision Making: Novel approaches to trajectory situation awareness and weapon target assignment in adversarial environments demonstrated the field's move toward high-stakes, real-world deployment.

Limitations and Future Work

As a proceedings volume, the limitations are generally specific to the individual papers. However, the editors note that the field is still evolving in its ability to handle "complex problems that would otherwise be difficult or impossible for humans to solve alone."

Future work suggested by the scope of the papers includes:

  • Improving the interpretability of cognitive models.
  • Scaling UAV swarm situation awareness for more complex environments.
  • Integrating large language models (LLMs) more deeply into robotic task assignment and decision-making frameworks.

Summary updated

BibTeX
@inproceedings{sun2024cognitive,
  author = {Fuchun Sun and Jianmin Li},
  title = {{Cognitive Computation and Systems}},
  booktitle = {Communications},
  year = {2024},
  doi = {10.1007/978-981-97-0885-7},
}

Loading related content...

Figure