Artificial Intelligence for Space Professionals

Each attendee receives extensive notes, reference materials and a Certificate of Completion.

The course is for engineers, scientists, business managers, and engineering managers of diverse background and with varying levels of experience, including those who are new or familiar to artificial intelligence. This class would appeal to individuals who are involved in planning space projects, designing systems, building, testing, and during the operations of autonomous “robotic” or manned spacecraft.

1...Introduction.

Overview and background of Artificial Intelligence: logicism, neuroscience, psychology, and

computer science. Includes definitions, comparison between brain and computer, and a short

history of AI. Introduce the topics of intelligent agent, uncertainty, and probability. Touch on the Deliberative Paradigm.

2...AI in Space.

Where AI can be applied to space systems, starting from the design process and through

operations in space. AI is applied to requirements (both developing requirements or AI

requirements), analysis modeling, integration with hardware, and testing, especially analyzing

test data. This includes a detailed definition of autonomy and automatic systems.

3...AI in Space Use Case Studies.

There are three diverse case studies: rockets, rovers, and spacecraft. Rocket design and

operations, rover design and autonomous navigation, and spacecraft optimization and

autonomous operations in light of time delays.

4...Space AI Examples.

Deep Space 1’s Remote Agent experiment, Orion’s Vehicle System Management (mission

planning, resource management, health management, fault detection/isolation/recovery, scripting, caution & warning, abort logic, re-init/re-start), mission planning on XSS-11, and pose estimation for docking or berthing using artificial targets.

5...Thinking Machines.

Two fundamental approaches: symbolic logic or connectionist with parallel distributed

processing. Pros and Cons as they relate to weak AI technologies

6...Logical Approach.

Symbolic approach starts with models to represent knowledge. Logic is used for deductive

reasoning, problem solving, search, and planning based on knowledge representation and search trees.

7...Machine Learning Approach.

All AI systems have a component of learning, starting with the neuron in the perceptron, neural networks, and progressing to deep learning (having more than three layers). Generative AI is AI that is capable of generating text, images, and other media using generative models. Of particular interest are foundational or Large Language Models (LLMs) for realistic

conversations.

8...Gen AI Applications.

Using Generative Adversarial Networks (GANs), AI can generate Deep Fakes. AI has reached

parity in Natural Language Processing, with the best examples being ChatBots such as ChatGPT.

9...AI Hardware.

Processors are getting faster, while following Moore’s Law. Computers still are

constructed following the Von Neumann architecture but are getting broader based on Graphical Processing Units and more recently with Google’s Tensor Processing Units (specialized ASICs or FPGA chips).

10...Computer Vision.

Computer vision has two diametrically opposing approaches: bottoms-up starting with features

and observing the “whole” scene. Features are detected with detectors and filters but is slow and cumbersome. New learning techniques have re-energized computer vision to identify objects, including vertical features and moving objects.

11...AI Impact with Space Systems.

Issues with complexity, integration of software with hardware, and symbolic model checking are discussed. Complexity (vehicle or platform) has to deal with safety, reliability, and risk. This includes issues with data, displacement of jobs, and AI running amok.

12...Miscellaneous AI.

The Turing Test is often referred to as the imitation game and is a general test for achieving

human-level intelligence. This test is refuted with Searle’s Chinese Room scenario. There are

multiple issues with ethics. And there are potential issues that require addressing: consciousness and sentience.

13...Summary and Re-cap.

Time permitting, AI has many other weak-AI technologies that deserve mentioning: cybernetics, data mining, anomaly detection, pattern recognition, diagnostics, prognostics, combinatorial explosions, decision-making, expert systems, common sense, and robotics.