The Banū Mūsā and Mechanical Procedure: Ingenious Devices, Automata, and Embodied Algorithmic Reasoning

Last Updated June 22, 2026

The Banū Mūsā and Mechanical Procedure examines how mechanical devices can embody step-by-step reasoning. In the ninth-century Abbasid intellectual world, the Banū Mūsā brothers are associated with the Kitāb al-Ḥiyal, often translated as the Book of Ingenious Devices, a work describing a wide range of mechanical, hydraulic, pneumatic, automatic, and trick devices. The importance of these devices for algorithmic history is not only that they move. It is that they encode procedure in material form.

A mechanical device can be read as a physical algorithm. Inputs are applied. Fluids move. Pressure changes. Valves open and close. Floats rise. Compartments fill. Timing sequences unfold. Hidden constraints regulate visible behavior. The device produces an output because its parts are arranged in a rule-governed sequence.

This article treats Banū Mūsā mechanical design as procedural reasoning made tangible. It does not claim that the brothers built modern computers, robots, or software. Instead, it asks a more historically careful question: how did premodern mechanical engineers use sequences, thresholds, feedback, concealment, pressure, balance, and conditional action to make devices behave in patterned ways?

This article introduces the Banū Mūsā brothers, the Book of Ingenious Devices, hydraulic automata, pneumatic control, vessels, fountains, valves, siphons, floats, feedback, sequencing, thresholds, concealed mechanisms, mechanical explanation, procedural engineering, Islamic-world technology, manuscript diagrams, and the history of algorithmic reasoning in material systems. It argues that mechanical procedure belongs in this series because it shows how computation-like behavior can be embodied in physical arrangement before digital machines.

Why the Banū Mūsā Matter

The Banū Mūsā matter because their mechanical work shows procedural intelligence outside of written calculation alone. Algorithms are often imagined as abstract instructions, but procedures can also be built into matter. A device can make decisions in a limited sense when its structure creates different outcomes under different physical conditions.

The Book of Ingenious Devices is often described as containing about one hundred mechanical devices, many involving hydraulics, pneumatics, automatic behavior, and cleverly hidden control. Its historical importance lies in the way it joins practical mechanics, mathematical imagination, demonstration, entertainment, utility, and explanation.

The Banū Mūsā show that procedure can be engineered, not only written.

Back to top ↑

Mechanical Procedure as Algorithmic Reasoning

A mechanical procedure is an ordered arrangement of physical causes. It begins with an initial condition: a vessel is filled, a passage is blocked, a float is positioned, a siphon is primed, or a valve is set. Then the device evolves according to its structure. Liquid rises. Pressure changes. A hidden passage opens. A flow switches direction. The visible action appears surprising, but the underlying process is rule-governed.

This resembles algorithmic reasoning because the outcome depends on steps, states, constraints, and transitions. The device does not “think” like a person. It does not compute in the digital sense. But it does embody a designed sequence that transforms inputs into outputs.

Mechanical procedure is algorithmic when the device’s physical arrangement determines a repeatable sequence of behavior.

Back to top ↑

The Book of Ingenious Devices

The Book of Ingenious Devices is famous because it documents mechanisms rather than merely celebrating marvels. Its diagrams and descriptions make devices intelligible enough to analyze. The work includes vessels, fountains, lamps, automatic controls, and other mechanisms that rely on fluid behavior, pressure, concealed channels, and sequential operations.

The word “ingenious” is important. These devices were not only useful machines in the modern industrial sense. Many were demonstrations, courtly objects, teaching artifacts, and displays of technical mastery. A vessel that dispenses a measured quantity, a fountain that changes pattern, or a lamp that regulates itself can be both entertaining and intellectually serious.

The book matters because it makes mechanical reasoning visible as a transmissible procedure.

Back to top ↑

Hydraulics, Pneumatics, and Material Logic

Many Banū Mūsā devices depend on hydraulics and pneumatics. Hydraulics uses the behavior of liquids. Pneumatics uses air and pressure. Together, they create a material logic: a chamber fills, air is compressed, pressure drives fluid, a siphon starts, a hidden passage transfers force, or a valve changes direction.

This material logic is procedural because each physical condition enables the next. The “program” is not stored in text after construction. It is stored in geometry, channel size, chamber placement, fluid level, valve shape, and pressure relation.

Hydraulic and pneumatic design shows how physical systems can encode rules without symbols.

Back to top ↑

Valves, Floats, and Thresholds

Valves and floats are especially important for mechanical procedure because they introduce conditional behavior. A valve can block one route and open another. A float can rise with fluid level and trigger a change. A threshold determines when the next step occurs.

This makes the device more than a passive container. It becomes a state machine in material form. It behaves differently depending on internal conditions. The action is not arbitrary. It follows from designed constraints.

A valve is not simply a part. In mechanical procedure, it is a decision point.

Back to top ↑

Sequencing, Timing, and Hidden Control

Many ingenious devices depend on the timing of events. A visible action may happen only after an invisible chamber fills. A fountain may change pattern after pressure shifts. A vessel may pour only a measured amount because hidden channels regulate how long flow continues.

Timing can be created by flow rate, reservoir size, channel width, pressure relation, or float motion. This is mechanical sequencing. The device moves through a designed series of states, and each state prepares the next.

Mechanical timing is procedure stretched into physical time.

Back to top ↑

Feedback and Self-Regulation

Some mechanical devices regulate themselves by letting the output or state of the system affect subsequent behavior. A float responds to liquid level. A lamp may maintain oil supply. A vessel may stop pouring when internal conditions change. These are not feedback systems in the full modern control-theory sense, but they show an important premodern intuition: a device can monitor a condition through its own structure and adjust behavior accordingly.

Feedback is central to algorithmic reasoning because it connects output, state, and future action. Instead of a single straight-line sequence, the system responds to its own condition.

Self-regulation shows how mechanical systems can contain a primitive logic of adjustment.

Back to top ↑

Trick Vessels and Controlled Surprise

The Banū Mūsā devices often include vessels that behave unexpectedly: pouring one liquid but not another, dispensing measured quantities, changing flow, or producing effects that conceal their mechanism. These “tricks” should not be dismissed as mere entertainment. Controlled surprise can be an engineering demonstration. It reveals mastery over hidden channels, pressure, flow, and timing.

A trick vessel is algorithmic because it separates appearance from internal procedure. The observer sees a surprising output. The engineer understands a rule-governed internal process. This distinction between interface and mechanism is central to later computing systems as well.

A mechanical trick is not anti-scientific; it can be a demonstration of procedural control.

Back to top ↑

Fountains and Alternating Behavior

Fountains are especially important in the history of mechanical procedure because they make hidden dynamics visible. Water moves through channels, pressure changes, outlets alternate, and patterns appear in time. A fountain can become a display of controlled behavior: one pattern, then another, then another.

Alternating fountains demonstrate state change. The visible output depends on internal configuration. If the device cycles, it shows repeated sequence. If the fountain changes based on pressure or flow, it shows condition-dependent behavior.

A fountain can be read as a visible program of flow.

Back to top ↑

Diagrams, Instructions, and Reproducibility

The procedural importance of the Banū Mūsā is strengthened by the presence of diagrams and instructions. A device described only as a wonder remains hard to analyze. A device described through parts, channels, chambers, and sequences becomes reproducible. Documentation transforms mechanical skill into transmissible knowledge.

This is why the Book of Ingenious Devices belongs beside mathematical and astronomical texts in the history of computation. It is not just a catalogue. It is a technical communication system. It tells readers how hidden structure produces visible action.

Documentation is what turns a clever object into engineering knowledge.

Back to top ↑

Engineering as Theory in Action

Mechanical procedure joins theory and practice. A device works only if physical principles are understood well enough to control them. Pressure, flow, balance, volume, geometry, timing, and material construction must all be coordinated. Engineering here is not merely craft; it is reasoning through matter.

This matters for the history of algorithms because it challenges a narrow view of computation as only written symbols. The Banū Mūsā show that systematic reasoning can be embodied in a device. Procedure can be physical. Explanation can be diagrammatic. Verification can be operational: the device either performs as designed or it does not.

Mechanical engineering is theory made accountable to performance.

Back to top ↑

From Mechanical Procedure to Automation

It is tempting to call Banū Mūsā devices robots or programmable machines. That language can be evocative, but it needs care. These devices do show automatic behavior, conditional response, and sequential control. They do not have general-purpose programmability in the modern computational sense.

A better phrase is mechanical procedure. This captures the essential point without anachronism. The devices embody rules. They transform inputs into outputs. They use hidden state and material conditions. They can repeat designed behavior. They belong to the history of automation, but not because they are digital computers. They belong because they show how action can be delegated to structure.

The Banū Mūsā belong to automation history, but their importance is best understood through mechanical procedure.

Back to top ↑

Origin Stories and Careful Interpretation

The Banū Mūsā should not be used to tell a simplistic origin story in which “robots” suddenly appear in ninth-century Baghdad. Mechanical devices existed in earlier Greek, Hellenistic, Persian, Indian, Chinese, and other traditions, and the Islamic-world engineering tradition received, transformed, extended, and documented multiple inheritances. The important point is not isolated invention. It is transmission, synthesis, design, and procedural refinement.

Careful interpretation also avoids dismissing these devices as mere toys. Courtly, playful, and demonstrative devices can still contain serious engineering. A trick vessel can teach pressure and flow. A fountain can demonstrate sequencing. A self-regulating lamp can reveal feedback. The context may be aesthetic or entertaining, but the mechanism is analytical.

A careful history treats Banū Mūsā mechanics as material reasoning, not myth.

Back to top ↑

Back to top ↑

Mathematics, Computation, and Modeling

A mechanical device can be modeled as a state transition system:

A threshold mechanism can be represented as:

A flow-rate delay can be represented simply as:

A mechanical procedure can be summarized as:

These formulas use modern notation to make the computational structure visible. They are interpretive models, not claims that the Banū Mūsā used this symbolic notation.

Back to top ↑

Python Workflow: Mechanical Procedure Map

The Python workflow below creates a dependency-light interpretive map of Banū Mūsā mechanical procedure. It scores themes by mechanical structure, procedural sequence, conditional control, hidden state, feedback potential, historical significance, ethical caution, and modern resonance, then writes reproducible CSV and JSON outputs.

# banu_musa_mechanical_procedure_map.py
# Dependency-light workflow for mapping mechanical devices as embodied procedure.

from __future__ import annotations

from dataclasses import asdict, dataclass
from pathlib import Path
from statistics import mean
import csv
import json
from datetime import datetime, timezone

ARTICLE_ROOT = Path(__file__).resolve().parents[1]
TABLES = ARTICLE_ROOT / "outputs" / "tables"
JSON_DIR = ARTICLE_ROOT / "outputs" / "json"


@dataclass(frozen=True)
class MechanicalProcedureConfig:
    article: str = "banu_musa_and_mechanical_procedure"
    core_threshold: float = 0.80
    high_control_threshold: float = 0.86


def timestamp_utc() -> str:
    return datetime.now(timezone.utc).isoformat()


def write_csv(path: Path, rows: list[dict[str, object]]) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    if not rows:
        path.write_text("", encoding="utf-8")
        return
    fieldnames = sorted({key for row in rows for key in row.keys()})
    with path.open("w", newline="", encoding="utf-8") as handle:
        writer = csv.DictWriter(handle, fieldnames=fieldnames, extrasaction="ignore")
        writer.writeheader()
        writer.writerows(rows)


def write_json(path: Path, payload: object) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    path.write_text(json.dumps(payload, indent=2, sort_keys=True), encoding="utf-8")


def mechanical_procedure_themes() -> list[dict[str, object]]:
    return [
        {"theme_id": "hydraulic_material_logic", "mechanical_structure": 0.96, "procedural_sequence": 0.90, "conditional_control": 0.88, "hidden_state": 0.90, "feedback_potential": 0.84, "historical_significance": 0.94, "ethical_caution": 0.82, "modern_resonance": 0.94},
        {"theme_id": "pneumatic_pressure_control", "mechanical_structure": 0.94, "procedural_sequence": 0.88, "conditional_control": 0.90, "hidden_state": 0.92, "feedback_potential": 0.84, "historical_significance": 0.92, "ethical_caution": 0.82, "modern_resonance": 0.92},
        {"theme_id": "valves_floats_thresholds", "mechanical_structure": 0.96, "procedural_sequence": 0.92, "conditional_control": 0.98, "hidden_state": 0.88, "feedback_potential": 0.92, "historical_significance": 0.94, "ethical_caution": 0.82, "modern_resonance": 0.96},
        {"theme_id": "sequencing_and_timing", "mechanical_structure": 0.92, "procedural_sequence": 0.98, "conditional_control": 0.92, "hidden_state": 0.90, "feedback_potential": 0.86, "historical_significance": 0.92, "ethical_caution": 0.82, "modern_resonance": 0.96},
        {"theme_id": "trick_vessels_interface_mechanism", "mechanical_structure": 0.90, "procedural_sequence": 0.90, "conditional_control": 0.92, "hidden_state": 0.96, "feedback_potential": 0.82, "historical_significance": 0.90, "ethical_caution": 0.84, "modern_resonance": 0.94},
        {"theme_id": "fountains_and_alternating_behavior", "mechanical_structure": 0.94, "procedural_sequence": 0.94, "conditional_control": 0.90, "hidden_state": 0.88, "feedback_potential": 0.88, "historical_significance": 0.92, "ethical_caution": 0.82, "modern_resonance": 0.94},
        {"theme_id": "diagrams_and_reproducibility", "mechanical_structure": 0.88, "procedural_sequence": 0.92, "conditional_control": 0.86, "hidden_state": 0.86, "feedback_potential": 0.80, "historical_significance": 0.96, "ethical_caution": 0.84, "modern_resonance": 0.94},
    ]


def score_theme(row: dict[str, object], config: MechanicalProcedureConfig) -> dict[str, object]:
    mechanical_score = mean([
        float(row["mechanical_structure"]),
        float(row["procedural_sequence"]),
        float(row["conditional_control"]),
        float(row["hidden_state"]),
        float(row["feedback_potential"]),
        float(row["historical_significance"]),
        float(row["ethical_caution"]),
        float(row["modern_resonance"]),
    ])

    if mechanical_score >= config.core_threshold and float(row["conditional_control"]) >= config.high_control_threshold:
        interpretive_status = "core_mechanical_procedure_thread"
    elif mechanical_score >= config.core_threshold:
        interpretive_status = "major_mechanical_procedure_thread"
    else:
        interpretive_status = "supporting_mechanical_procedure_thread"

    return {
        "theme_id": row["theme_id"],
        "mechanical_structure": round(float(row["mechanical_structure"]), 6),
        "procedural_sequence": round(float(row["procedural_sequence"]), 6),
        "conditional_control": round(float(row["conditional_control"]), 6),
        "hidden_state": round(float(row["hidden_state"]), 6),
        "feedback_potential": round(float(row["feedback_potential"]), 6),
        "historical_significance": round(float(row["historical_significance"]), 6),
        "ethical_caution": round(float(row["ethical_caution"]), 6),
        "modern_resonance": round(float(row["modern_resonance"]), 6),
        "mechanical_score": round(mechanical_score, 6),
        "interpretive_status": interpretive_status,
    }


def interpretation_cautions() -> list[dict[str, str]]:
    return [
        {"caution": "do_not_call_the_devices_modern_robots", "meaning": "The devices show automatic behavior and mechanical procedure, not modern robotics or AI."},
        {"caution": "do_not_dismiss_trick_devices_as_toys", "meaning": "Courtly or surprising devices can still demonstrate serious hydraulic and pneumatic reasoning."},
        {"caution": "do_not_treat_mechanics_as_non_computational", "meaning": "Mechanical systems can embody rule-governed state transitions and conditional behavior."},
        {"caution": "do_not_ignore_material_constraints", "meaning": "Flow, pressure, leakage, friction, and construction determine whether procedure works."},
        {"caution": "do_not_create_single_origin_myths", "meaning": "Study Greek, Hellenistic, Islamic-world, and later engineering traditions as transmission and transformation."},
    ]


def main() -> None:
    config = MechanicalProcedureConfig()
    themes = mechanical_procedure_themes()
    scored = [score_theme(row, config) for row in themes]
    cautions = interpretation_cautions()

    summary = {
        "article": config.article,
        "timestamp_utc": timestamp_utc(),
        "themes_reviewed": len(scored),
        "core_threads": sum(1 for row in scored if row["interpretive_status"] == "core_mechanical_procedure_thread"),
        "major_threads": sum(1 for row in scored if row["interpretive_status"] == "major_mechanical_procedure_thread"),
        "supporting_threads": sum(1 for row in scored if row["interpretive_status"] == "supporting_mechanical_procedure_thread"),
        "mean_mechanical_score": round(mean(float(row["mechanical_score"]) for row in scored), 6),
        "cautions": len(cautions),
        "interpretation": "Banū Mūsā mechanical devices should be studied as embodied procedure: input, state, condition, transition, output, timing, feedback, and material constraint.",
    }

    write_csv(TABLES / "mechanical_procedure_themes.csv", themes)
    write_csv(TABLES / "mechanical_procedure_map.csv", scored)
    write_csv(TABLES / "interpretation_cautions.csv", cautions)
    write_csv(TABLES / "mechanical_procedure_summary.csv", [summary])

    write_json(JSON_DIR / "mechanical_procedure_config.json", asdict(config))
    write_json(JSON_DIR / "mechanical_procedure_map.json", scored)
    write_json(JSON_DIR / "interpretation_cautions.json", cautions)
    write_json(JSON_DIR / "mechanical_procedure_summary.json", summary)

    print("Banū Mūsā mechanical procedure map complete.")
    print(TABLES / "mechanical_procedure_summary.csv")


if __name__ == "__main__":
    main()

This workflow turns mechanical procedure into a reproducible interpretive artifact: hydraulic logic, pneumatic control, valves, floats, thresholds, sequencing, hidden state, feedback, diagrams, historical significance, and interpretive caution are documented together.

Back to top ↑

R Workflow: Mechanical Procedure Diagnostics

The R workflow reads the generated CSV outputs, summarizes mechanical-procedure themes, visualizes theme dimensions, and writes an additional diagnostic table.

# banu_musa_mechanical_procedure_summary.R
args <- commandArgs(trailingOnly = FALSE)
file_arg <- grep("^--file=", args, value = TRUE)

if (length(file_arg) > 0) {
  script_path <- normalizePath(sub("^--file=", "", file_arg[1]), mustWork = TRUE)
  article_root <- normalizePath(file.path(dirname(script_path), ".."), mustWork = TRUE)
} else {
  article_root <- getwd()
}

setwd(article_root)

tables_dir <- file.path(article_root, "outputs", "tables")
figures_dir <- file.path(article_root, "outputs", "figures")
dir.create(tables_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(figures_dir, recursive = TRUE, showWarnings = FALSE)

map_path <- file.path(tables_dir, "mechanical_procedure_map.csv")
summary_path <- file.path(tables_dir, "mechanical_procedure_summary.csv")

if (!file.exists(map_path)) {
  stop(paste("Missing", map_path, "Run the Python workflow first."))
}

mechanical_map <- read.csv(map_path, stringsAsFactors = FALSE)
summary <- read.csv(summary_path, stringsAsFactors = FALSE)

png(file.path(figures_dir, "mechanical_procedure_dimensions.png"), width = 1200, height = 850)
score_matrix <- t(as.matrix(mechanical_map[, c("mechanical_structure", "procedural_sequence", "conditional_control", "hidden_state", "feedback_potential", "historical_significance", "ethical_caution", "modern_resonance")]))
barplot(score_matrix,
        beside = TRUE,
        names.arg = mechanical_map$theme_id,
        las = 2,
        ylim = c(0, 1),
        ylab = "Interpretive Score",
        main = "Banū Mūsā Mechanical Procedure Dimensions")
legend("bottomright",
       legend = rownames(score_matrix),
       cex = 0.72,
       bty = "n")
grid()
dev.off()

png(file.path(figures_dir, "mechanical_score_by_theme.png"), width = 1000, height = 750)
barplot(mechanical_map$mechanical_score,
        names.arg = mechanical_map$theme_id,
        las = 2,
        ylab = "Mechanical Procedure Score",
        main = "Mechanical Procedure Score by Theme")
grid()
dev.off()

r_summary <- data.frame(
  themes_reviewed = summary$themes_reviewed[1],
  core_threads = summary$core_threads[1],
  major_threads = summary$major_threads[1],
  supporting_threads = summary$supporting_threads[1],
  mean_mechanical_score = summary$mean_mechanical_score[1],
  cautions = summary$cautions[1],
  diagnostic_note = "Banū Mūsā mechanical devices should be studied as embodied procedure: input, state, condition, transition, output, timing, feedback, and material constraint."
)

write.csv(r_summary, file.path(tables_dir, "r_mechanical_procedure_diagnostic_summary.csv"), row.names = FALSE)
print(r_summary)

The R layer makes the interpretive structure visible: hydraulic logic, pneumatic pressure control, valves, floats, thresholds, trick vessels, alternating fountains, diagrams, feedback, and caution can be examined as related but distinct dimensions of mechanical procedure.

Back to top ↑

GitHub Repository

The companion repository contains reproducible workflows, synthetic interpretive data, outputs, calculators, documentation, and multilingual examples for this article.

Back to top ↑

A Practical Method for Studying Mechanical Procedure

A careful study of mechanical procedure should ask how a device transforms initial conditions into visible behavior. The method below treats machines as embodied procedures.

This method treats mechanical devices as procedures made from matter, geometry, and motion.

Back to top ↑

Common Pitfalls

The first pitfall is calling the Banū Mūsā devices modern robots. The second is dismissing the devices as toys. The third is treating physical mechanisms as unrelated to computation. The fourth is ignoring material constraints such as leakage, friction, pressure, and construction accuracy.

Mechanical procedure becomes clearer when historical machines are neither diminished nor exaggerated.

Back to top ↑

Why Mechanical Procedure Belongs in Algorithmic Reasoning

The Banū Mūsā and mechanical procedure belong in algorithmic reasoning because they show that procedure can be embodied. A device can contain inputs, states, thresholds, transitions, outputs, delays, hidden channels, and feedback. It can transform initial conditions into patterned behavior through physical design.

This history expands the meaning of computation. Computation is not only symbol manipulation, numerical calculation, software, or digital execution. It is also rule-governed transformation across media. In mechanical devices, rules are encoded in channels, chambers, valves, floats, pressures, and timing.

The lesson for modern systems is direct. Algorithms are never only abstract. They are implemented in materials, infrastructures, interfaces, and institutions. The Banū Mūsā remind us that procedure always has a body, whether that body is a manuscript diagram, a hydraulic device, a mechanical controller, a server farm, or an AI system. AI belongs in the toolkit, not in control.

Back to top ↑

Related Articles

• Al-Kindī, Frequency Analysis, and the Birth of Cryptanalysis
• Al-Jazarī, Automata, and Programmable Mechanisms
• Astronomical Tables, Calendars, and Algorithmic Prediction
• Algorithms in Systems Modeling
• AI Agents, Tool Use, and Procedural Autonomy

Back to top ↑

Further Reading

• Banū Mūsā (1979) The Book of Ingenious Devices. Translated and annotated by D.R. Hill. Dordrecht: D. Reidel.
• Hill, D.R. (1974) The Book of Knowledge of Ingenious Mechanical Devices. Dordrecht: D. Reidel.
• Hill, D.R. (1998) Studies in Medieval Islamic Technology. Aldershot: Ashgate.
• Encyclopaedia Iranica (n.d.) ‘Banū Mūsā’.
• Lemelson Center for the Study of Invention and Innovation (2020) ‘Ingenious Devices’. Smithsonian Institution.
• 1001 Inventions (n.d.) ‘Banu Musa and the Science of Tricks’.
• Rosheim, M.E. (1994) Robot Evolution: The Development of Anthrobotics. New York: Wiley.

Back to top ↑

References

• Banū Mūsā (1979) The Book of Ingenious Devices. Translated and annotated by D.R. Hill. Dordrecht: D. Reidel.
• Encyclopaedia Iranica (n.d.) ‘Banū Mūsā’. Available at: https://www.iranicaonline.org/articles/banu-musa-the-name-applied-to-three-brothers-abbasid-astronomers-whose-father-was-musa-b/.
• Hill, D.R. (1974) The Book of Knowledge of Ingenious Mechanical Devices. Dordrecht: D. Reidel.
• Hill, D.R. (1998) Studies in Medieval Islamic Technology. Aldershot: Ashgate.
• Lemelson Center for the Study of Invention and Innovation (2020) ‘Ingenious Devices’. Smithsonian Institution. Available at: https://invention.si.edu/invention-stories/ingenious-devices.
• Rosheim, M.E. (1994) Robot Evolution: The Development of Anthrobotics. New York: Wiley.
• 1001 Inventions (n.d.) ‘Banu Musa and the Science of Tricks’. Available at: https://www.1001inventions.com/banu-musa/.

Back to top ↑