Getting into robotics as a software engineer

Original Article ↗ Hacker News Discussion ↗

Software engineers curious about robotics weigh practical entry paths, from starting with Arduinos, Raspberry Pis, and low-cost drones to using simulators like CoppeliaSim and Isaac Sim before investing in hardware. Commenters stress that beyond programming in C++ or Python, real-world robotics work demands basic electronics skills, control theory (e.g., PID), and often familiarity with ROS, as well as an appreciation of the slower, less "agile" iteration cycles imposed by hardware. Many note a trade-off between the intellectual appeal and tangible impact of robotics and the typically higher pay and easier hiring pipelines in web, adtech, and fintech roles.

Entry paths & starter platforms

  • Many suggest starting from software with basic hardware: learn schematics, datasheets, oscilloscopes, and logic analyzers.
  • For home projects, Arduino is widely recommended over Raspberry Pi for beginners: simpler, no OS, quick upload–run cycle; cheap clones are fine.
  • Others list Raspberry Pi, Teensy, ESP32, Micro:BIT, plus steppers, servos, and sensors; ~$200 can stock a serious starter lab.
  • Caution that Raspberry Pi’s SD cards corrupt easily under power loss during debugging; some suggest simulators (e.g., Wokwi) and microcontrollers first.

Embedded complexity & debugging

  • Knowing C/C++ and using a scope is seen as valuable but insufficient.
  • Experienced voices stress that real difficulty is knowing where to probe, what to look for, and making edge cases reproducible under tight lab and budget constraints.
  • Real-world debugging involves emergent hardware issues (noise, timing, components out of spec) that hobby tinkering doesn’t fully prepare you for.

Math, control, and theory

  • Several software engineers feel underprepared in trig, geometry, and calculus when doing robotics/CAD/3D printing.
  • Some say basic trig/geometry is enough and many problems are solved in libraries.
  • Others argue strong trig and linear algebra, plus control theory, are crucial, especially for positioning, attitude control, and understanding matrices/quaternions; note that simple PIDs often replace overcomplicated custom logic.

Process & “agile” in hardware

  • Multiple comments claim agile, as practiced in software, clashes with hardware realities: PCB spins take days–weeks, and each iteration is expensive.
  • Strategies discussed: overbuilding prototypes with extra sensors/actuators, using dev kits and modular boards, leaving unpopulated footprints for future components.
  • Some are openly hostile to importing web-style agile/scrum, saying it has damaged hardware/robotics teams.

Careers, salaries, and paths in robotics

  • Strong perception that hardware/robotics generally pays less than adtech/fintech/web, which pulls many developers away.
  • Robotics is described as niche, dominated by early-stage startups with average salaries; high pay mostly at a few large firms.
  • Suggested entry path for software devs:
    • Learn C++ and/or Python plus embedded basics (RTOS, ISRs, buses like CAN/I²C/SPI).
    • Learn ROS, do tutorials and simulations, wire up a basic teleoperated robot.
    • Contribute to ROS/open-source packages and follow robotics job boards and conferences.

ROS and tooling

  • ROS 1 Noetic on Ubuntu is viewed as mature and stable.
  • ROS 2 is seen as promising for distributed/multi-robot systems but criticized for instability across versions, middleware quirks, and poor Python-node performance; rapid Ubuntu+ROS upgrade cycles are a pain.
  • Debate whether final products use ROS: some say it’s primarily academic; others note many commercial stacks either use ROS or reimplement similar messaging, parameters, and logging.

Drones, simulators, and specific platforms

  • For programmable drones, suggestions include Crazyflie (open hardware/firmware/software) and small platforms like Ryze Tello, though ease of development is questioned.
  • Isaac Sim is praised for simulation, especially with Jetson-based robots; some recommend more capable chassis (e.g., WaveRover) over older platforms like Jetbot.
  • Simulators such as CoppeliaSim are recommended to learn robotics without hardware investment.

Broader reflections on impact

  • Some reject the idea that “apps can’t solve real-world problems,” arguing software is critical to large-scale resource and people management.
  • Others lament that the best-paying work is often ad-tech or similar, seen as low social value but highly lucrative.
  • There’s strong enthusiasm that now is a particularly good time for robotics: better batteries, motors, radios, cameras, and cheap LIDAR make serious projects far more accessible than in past decades.