CORES v0.1

A robot hasone battery,one CPU, andone chanceto get home.

CORES decides which cognitive modules run, in what order, and under what constraints. Every cycle. Deterministically.

It is not the robot's brain. It is the infrastructure that keeps the brain from burning out.

Read the architecture decisions

Runtime Status

Live
84%
Battery
5
Modules Active
<1ms
Latency
128
Events

Mission Utility

Lexicographic
Cycle 0Cycle 80
Safety
100%
Energy
60%
Architecture

The stack is the interface.

Modules compose vertically. Each layer adds intelligence without breaking determinism. The runtime never knows which modules are loaded.

04

Lexicographic Scheduler

Multi-objective dispatch across safety, mission, and energy dimensions. Produces the one Pareto-optimal execution plan for each cycle.

03

Risk-Aware Knapsack

Energy-budgeted task selection modeled as a knapsack problem. Maximizes mission value within the available energy envelope.

02

Criticality Scheduling

Safety-critical modules receive bounded, non-preemptible execution slots with SIL-4 equivalent fault containment.

01

Runtime Foundation

Deterministic priority scheduler with O(1) dispatch and zero-allocation hot paths. Fixed-priority preemptive scheduling.

CRCORES Runtime
Deterministic cycle · 128 μs avg
Scheduling

One robot. Four policies. Four outcomes.

Each policy implements the same trait. Swap it at construction. The runtime never knows the difference.

Baseline

Priority

01
Navigation
P1
Locomotion
P2
StateEstimation
P3
Perception
P4
Mapping
P5
Safety
60
Mission
85
Energy
30
Safety-First

Criticality

02
StateEstimation
SIL-4
Navigation
SIL-3
Locomotion
SIL-3
Perception
SIL-2
Mapping
SIL-1
Safety
100
Mission
60
Energy
25
Optimized

Knapsack

03
Navigation
36%
Locomotion
24%
StateEstimation
18%
Perception
12%
Mapping
10%
Safety
75
Mission
90
Energy
85
Pareto-Optimal

Lexicographic

04
StateEstimation
S1
Navigation
S1
Locomotion
M2
Mapping
M3
Perception
E4
Safety
100
Mission
95
Energy
70
Test Suite

Failure is the input.

Twenty scenario families. Deterministic seeds. Every cycle is a new failure mode vector.

Hand-picked scenarios do not generalize. This generator spans every subsystem, then combines them. The same code drives 1000+ trial Monte Carlo workloads.

PowerThermalCommsSensorActuatorTiming
1Nominal Exploration
2Low Battery (5%)
3Obstacle Detected
4System Emergency
5Budget Exhaustion (30%)
6Sensor Failure (GPS+Camera)
7High Temperature (65°C)
8Communication Loss
9Navigation Loss (GPS denied)
10Camera Degraded
11GPS Drift
12LIDAR Failure
13Deadline Overload
14Memory Pressure
15Multi-Sensor Failure
16Thermal Throttling
17Network Partition
18Actuator Degradation
19Mission Change
20Unknown Environment

Monte Carlo Mode

The same ScenarioGenerator feeds 1000+ trial Monte Carlo workloads. Seeded RNG guarantees exact reproducibility.

1000+
Trials
100%
Reproducible
20
Seeds
Telemetry

Every cycle, measured.

Deterministic execution produces deterministic traces. Every tick is logged, measured, and accountable.

Tasks Scheduled+4.2%
12,847
Avg Latency-3.1%
128 μs
Safety Violations0%
0
Energy Efficiency+2.5%
87%

Runtime Cycle Trace

81 ticks
0
10
20
30
40
50
60
70
80
Active modulesBatteryCritical event