Olympus Station Emergency

🎯 Lab Objectives

In this simulation, you will experience the role of an AI Mission Support Agent responding to a critical life-support failure at a crewed Mars habitat. You will:

• Design the agent's reasoning process by making strategic decisions.
• Experience the constraints of communication delays and limited resources.
• Evaluate how different agent strategies lead to different mission outcomes.
• Build a working agent prototype using JavaScript and the Gemini API.

---

📡 Mission Telemetry Dashboard

Current station readings at Sol 247, 14:32 LMST (Local Mars Solar Time):

---

🔴 Phase 1: Emergency Response Simulation

You are the Olympus Station AI Agent. The CO2 scrubber has dropped to 72% efficiency. You cannot wait 28 minutes (round-trip) for Earth to respond. Walk through the following decision tree to see how agent choices affect outcomes.

---

✏️ Phase 2: Agent Design Exercise

Now that you have experienced the agent's decision-making process, design your own agent for a space scenario. Work in pairs and complete the following Agent Design Canvas:

---

💻 Phase 3: Code Implementation

Build a working agent prototype using JavaScript and the Gemini API. The agent will interact with a simulated telemetry system and make autonomous decisions.

Step 1: Project Setup

Create a new folder for your agent project and set up the following files:

mars-agent/
├── index.html          # Agent UI and console
├── agent.js            # Agent logic and orchestration loop
├── tools.js            # Tool definitions (sensors, alerts, etc.)
└── style.css           # Styling (use the course design system)

Step 2: Define the Agent Tools

Create tools.js with simulated Mars habitat tools:

// tools.js - Simulated Mars Habitat Tool Suite

const habitatTools = {
    readSensor: function(sensorId) {
        const sensors = {
            co2_scrubber: { value: 72, unit: "%", status: "DEGRADED",
                           trend: "declining", rate: "-1.5%/hr" },
            o2_reserve:   { value: 48, unit: "hours", status: "NOMINAL" },
            cabin_pressure: { value: 101.3, unit: "kPa", status: "NOMINAL" },
            temperature:  { value: 21.4, unit: "C", status: "NOMINAL" },
            radiation:    { value: 0.21, unit: "mSv/hr", status: "NOMINAL" }
        };
        return sensors[sensorId] || { error: "Unknown sensor: " + sensorId };
    },

    checkInventory: function(item) {
        const inventory = {
            co2_filter:     { quantity: 2, location: "Storage Bay C",
                              condition: "sealed" },
            o2_canister:    { quantity: 6, location: "Life Support Bay" },
            medical_kit:    { quantity: 3, location: "Med Bay Alpha" },
            repair_toolkit: { quantity: 1, location: "Engineering Bay" }
        };
        return inventory[item] || { error: "Item not found: " + item };
    },

    getCrewStatus: function() {
        return [
            { name: "Commander Chen", location: "Hab Module A",
              status: "Active", heartRate: 72 },
            { name: "Dr. Okafor",     location: "Med Bay",
              status: "Active", heartRate: 68 },
            { name: "Eng. Petrov",    location: "Engineering",
              status: "Active", heartRate: 75 },
            { name: "Sci. Nakamura",  location: "EVA - Exterior",
              status: "EVA",    heartRate: 82, evaTimeRemaining: "2h 15m" }
        ];
    },

    sendAlert: function(level, message) {
        console.log("[ALERT:" + level + "] " + message);
        return {
            sent: true, level: level, timestamp: new Date().toISOString(),
            earthDeliveryETA: "14 minutes"
        };
    },

    queryMaintenanceLog: function(system) {
        const logs = {
            co2_scrubber: {
                lastService: "Sol 210", nextScheduled: "Sol 280",
                notes: "Filter replacement every 70 sols. Current filter installed Sol 210.",
                procedure: "1. Shut down scrubber unit. 2. Depressurize filter housing. " +
                           "3. Remove spent filter (caution: residue). " +
                           "4. Insert new filter (verify seal). " +
                           "5. Repressurize and restart. 6. Monitor for 30 min."
            }
        };
        return logs[system] || { error: "No logs for: " + system };
    }
};

export default habitatTools;

Step 3: Build the Agent Loop

Create agent.js with the core ReAct orchestration loop:

// agent.js - Mars Mission Agent (ReAct Pattern)
import habitatTools from './tools.js';

const SYSTEM_PROMPT = `You are ARIA (Autonomous Response & Intelligence Agent),
the AI mission support system for Olympus Station on Mars.

PRIORITIES (in order):
1. Crew safety is paramount.
2. Maintain life support systems.
3. Preserve mission objectives.
4. Conserve resources.

CONSTRAINTS:
- Earth communication delay: 14 minutes one-way.
- You MUST act autonomously for time-critical decisions.
- Always explain your reasoning before acting.
- Escalate to crew commander for irreversible actions.

AVAILABLE TOOLS:
- readSensor(sensorId): Read environmental sensor data
- checkInventory(item): Check supply inventory
- getCrewStatus(): Get all crew member locations and vitals
- sendAlert(level, message): Send alert (levels: INFO, WARNING, CRITICAL)
- queryMaintenanceLog(system): Retrieve maintenance procedures

Respond in JSON: { "thought": "...", "action": "tool_name",
                    "params": {...} }
Or if done:      { "thought": "...", "answer": "..." }`;

class MarsAgent {
    constructor() {
        this.conversationHistory = [
            { role: "system", content: SYSTEM_PROMPT }
        ];
        this.maxSteps = 8;
        this.stepCount = 0;
    }

    async run(initialObservation) {
        this.log("system", "ARIA Agent initialized. Processing alert...");
        let observation = initialObservation;

        while (this.stepCount < this.maxSteps) {
            this.stepCount++;
            this.conversationHistory.push({
                role: "user",
                content: "Observation: " + observation
            });

            // Call LLM (replace with actual Gemini API call)
            const response = await this.think();

            if (response.answer) {
                this.log("agent", "Final Report: " + response.answer);
                return response.answer;
            }

            if (response.action) {
                this.log("agent", "Thought: " + response.thought);
                this.log("action", "Calling tool: " +
                    response.action + "(" +
                    JSON.stringify(response.params) + ")");

                const result = this.executeTool(
                    response.action, response.params
                );
                observation = JSON.stringify(result);
                this.log("info", "Result: " + observation);
            }
        }
        return "Max steps reached. Escalating to crew commander.";
    }

    executeTool(toolName, params) {
        if (habitatTools[toolName]) {
            return habitatTools[toolName](...Object.values(params));
        }
        return { error: "Unknown tool: " + toolName };
    }

    log(type, message) {
        const time = new Date().toLocaleTimeString();
        console.log("[" + time + "] [" + type.toUpperCase() + "] " + message);
        // In a real implementation, update the UI console here
    }

    async think() {
        // Placeholder: Replace with actual Gemini API call
        // const response = await fetch('https://generativelanguage.googleapis.com/...');
        // For the lab, use the simulated responses below
        return this.simulateResponse();
    }

    simulateResponse() {
        // Simulated agent responses for the lab exercise
        const responses = [
            {
                thought: "CO2 alert received. First, I need to assess all " +
                         "sensor data to understand the full picture.",
                action: "readSensor",
                params: { sensorId: "co2_scrubber" }
            },
            {
                thought: "Scrubber at 72% and declining. I need to check " +
                         "if we have spare filters.",
                action: "checkInventory",
                params: { item: "co2_filter" }
            },
            {
                thought: "2 spare filters available in Bay C. Let me check " +
                         "maintenance procedures.",
                action: "queryMaintenanceLog",
                params: { system: "co2_scrubber" }
            },
            {
                thought: "I have the procedure. Let me check crew " +
                         "availability before assigning tasks.",
                action: "getCrewStatus",
                params: {}
            },
            {
                thought: "3 crew in habitat, 1 on EVA. Sending situation " +
                         "report to Earth while we begin repairs.",
                action: "sendAlert",
                params: {
                    level: "WARNING",
                    message: "CO2 scrubber degraded 72%. Initiating " +
                             "autonomous filter replacement. ETA: 2 hours."
                }
            },
            {
                thought: "All information gathered. Generating action plan.",
                answer: "ACTION PLAN: 1) Commander Chen and Eng. Petrov to " +
                        "Storage Bay C for filter retrieval. 2) Dr. Okafor " +
                        "to monitor crew vitals and cabin atmosphere. " +
                        "3) Sci. Nakamura continue EVA (safe, 2h remaining). " +
                        "4) Earth notified, response ETA 28 min. " +
                        "5) Agent monitoring scrubber trend. If efficiency " +
                        "drops below 60%, will activate backup O2."
            }
        ];
        return responses[Math.min(this.stepCount - 1, responses.length - 1)];
    }
}

export default MarsAgent;

Step 4: Run and Extend

---

🖥️ Live Agent Console

Watch the ARIA agent respond to the Olympus Station emergency in real-time:

---