Skip to content

Releases

v85 · 2026-11-07

Choose a tag to compare

@github-actions github-actions released this 07 Nov 22:19
83e45c8
  • An open lesson's header is tidier: All lessons, the group and Copy link on one line, Previous and Next on the line
    below, Previous on the left and Next on the right. On a wide screen it all fits on one line; on a phone the buttons
    read "← Previous" and "Next →".
  • In the lesson list each lesson's minutes line up in a column at the right, and the "advanced" tag sits right after
    the title. When a title wraps onto two lines, the progress circle stays beside the lesson's number.

v84 · 2026-11-07

Choose a tag to compare

@github-actions github-actions released this 07 Nov 21:19
0346ddc
  • The page now opens in Basic: the map, the four parts of the robot (Localize, Perceive, Plan, Control) with their
    main settings, one plot and three metrics. Advanced (top right, or View → Show) brings back every tab, setting
    and plot, the event log and Runs, and the page remembers your choice. Nothing is out of reach in Basic: when a search
    result, a layout or a shared link needs something Basic hides, the page switches to Advanced and says so, and each
    Basic tab ends with how many more settings Advanced has.
  • Analyze is now Loop analysis, opened from the top of the Control tab ("Analyze this loop →") or with the A key.
    It says what it shows, has a "← Back to the simulator" button, and its side panel keeps only Control and Vehicle,
    the two tabs it reads. Lesson and chapter links to it work as before.
  • A new first group in Learn, Foundations: ten short lessons for an engineering undergraduate, one idea each (the
    loop, odometry, the Kalman filter, uncertainty, lidar mapping, the costmap, A*, pure pursuit, delay, people). Start
    here's first pass is now these ten.
  • Every lesson's derivations are folded under Show the math, so the text reads without them; open it when you want
    the formulas. Reading times no longer count the folded math. Each group says the level it is written at, and the
    lessons that assume the most are marked advanced in the list.
  • Buttons now come in three sizes that say what they do: one large filled button for the main action (Run; Start with
    lesson N in Learn), one size for ordinary actions, including each lesson's experiment buttons, and small buttons only
    inside dense rows such as the telemetry dock and the Runs list. Close (×) and other icon buttons are square and the
    same size everywhere, actions that change the lab (All on / All off beacons, Forget the previous run, Pin, Reset a
    lesson, Copy link) are buttons rather than underlined text, and on phones the controls are a little larger for touch.
  • While a lesson is open, the telemetry shows the event log and every metric even in Basic, because the lessons point at
    them; a Foundations lesson never switches the page to Advanced.

v83 · 2026-11-06

Choose a tag to compare

@github-actions github-actions released this 06 Nov 14:09
6ea5856
  • The page has an icon in the browser tab, and a link to the demo now comes with a preview card on sites and in chats
    that show one: the title, a line on what the page is, and a picture of the busy floor in dark mode during a run.

v82 · 2026-11-05

Choose a tag to compare

@github-actions github-actions released this 05 Nov 13:11
78d9aba
  • On a phone, the page no longer gets stuck at the bottom. After you scrolled down to change the floor, it could refuse
    to scroll back up to the map, most of all on iPhone and in the Claude app. The status strip and the metrics were
    redrawn several times a second even when nothing changed, and on iPhone each redraw nudged the page's height by a
    pixel, which kept pulling the page back to the bottom. They are now redrawn only when a value changes, and while a
    run is going the telemetry and the status strip keep the tallest height they reached (until you change floor, reset
    or fold the dock), so the page no longer shrinks under your finger as the numbers change.

v81 · 2026-11-04

Choose a tag to compare

@github-actions github-actions released this 04 Nov 18:51
17a25bc
  • On a phone, the level menu no longer leaves the page looking stuck. Swiping on the menu after picking a level used
    to scroll the page hidden behind it, so nothing on screen moved. Now a swipe only scrolls the menu itself, the menu
    has a × to close it, and its last lines are no longer hidden behind the bottom bar.

v80 · 2026-10-31

Choose a tag to compare

@github-actions github-actions released this 31 Oct 20:20
4a7514f
  • Chapter 4 has a new section, Speed planning: the ST graph (§6), on how the robot slows and waits for people
    without leaving its path: the grid of time against distance along the path, the cost of each cell near a predicted
    person, the search for the cheapest speed profile, when it gives up and asks for a new path, and a worked encounter
    in which a person turns back toward the robot. The later sections of the chapter move up by one.
  • Control → Response to agents now lists all four options with every controller and dims the ones it cannot use, with
    the reason when you hover or tap: with MPC, velocity obstacles and the reactive stop; with DWA or your own
    controller, everything but its own handling of people. Before, the row was missing for DWA and showed only two
    options for MPC.

v79 · 2026-10-31

Choose a tag to compare

@github-actions github-actions released this 31 Oct 19:34
82ed809
  • With the lidar switched off (Sensors) and the radar on, setting Perceive → Tracker inputs to Radar now tracks people
    from the radar alone, so speed planning and velocity obstacles can still give way to them. Before, the tracker
    waited for lidar scans that never came, and the robot drove straight through people. The radar only looks ahead,
    so someone coming from the side or behind is seen late or not at all. With the lidar off and Tracker inputs left
    on Lidar, the health panel and the Perceive tab now say how to switch.

v78 · 2026-10-31

Choose a tag to compare

@github-actions github-actions released this 31 Oct 19:13
a2a89d8
  • Every switch of options (Agent tracking, Map source, Simulate · Analyze · Learn, the plot and log pickers) now fills
    the option that is on with blue, so which one is chosen reads at a glance. With agent tracking off, the Perceive
    tab says so first and offers a button to turn on lidar tracking.

v77 · 2026-10-31

Choose a tag to compare

@github-actions github-actions released this 31 Oct 19:06
4b3f8e6
  • Custom level → Generate: the number fields now line up, each label beside its box in even columns, so the form
    reads as a table instead of a ragged run of fields.
  • The first-run hint gives the first pass's length as Start here counts it (99 minutes since the lessons were
    rewritten; it said 95).

v76 · 2026-10-31

Choose a tag to compare

@github-actions github-actions released this 31 Oct 17:15
b2c7b5a
  • The nine chapters are rewritten in plainer prose, as the lessons were in v72 and v75: long sentences split, the
    reason next to the fact, every term explained where it first appears, and a few skipped steps written out. No
    heading, equation, table, live value or measured number changed.
  • Corrections found on the way, in the chapters and the lessons that point at them: a late lidar scan is placed
    from an estimate that lags with it, so the walls smear rather than shift (perception lesson 4 and chapter 3 §7,
    measured headless); a diagonal space-time move takes one step, checked at both ends; velocity obstacles sample 77
    candidates; the chamfer distance errs on the unsafe side and the lethal margin covers it; a radar update leaves an
    ellipse stretched across the line of sight; pure pursuit's short-lookahead oscillation comes from the delays, not
    from lost damping; and the Butterworth damping of LQR holds for any weight ratio when only position is weighted.